A substation flood intelligent monitoring method and system

Abstract

The application discloses a kind of substation flood intelligent monitoring method and system, it is related to electric power system safety monitoring field.The method is by building the multi-dimensional influence factor library including the bottom elevation of discharge port and pipe network permeation loss, combined with rainfall forecast executes dynamic water balance deduction, deduce according to the dynamic correction soil infiltration capacity of early rainless days, and according to the effective pressure difference of water level and the bottom of discharge port calculates net drainage capacity;Based on the deduction result, water level height field is generated by combining three-dimensional digital model for waterlogging overflow distribution;According to water level height field and equipment elevation, determine the flooding risk, and generate dispatching instruction by taking water center as target and combining prohibited operation area to optimize equipment layout coordinates;Through man-machine interaction correction parameter and trigger GPU acceleration recalculation to realize model online evolution.The application realizes the physical mechanism simulation and space intelligent scheduling under the microscale of substation, changes passive alarm into active risk deduction.

Description

Technical Field

[0001] This application belongs to the interdisciplinary field of artificial intelligence and power system safety monitoring, specifically relating to a method and system for intelligent monitoring of floods in substations. Background Technology

[0002] In recent years, flood monitoring technology for substations has mainly relied on water level sensors and meteorological data, issuing risk warnings by setting fixed thresholds. Building upon this, some existing technologies have further integrated geographic information systems and equipment ledgers, achieving preliminary assessments of the flooded area of ​​power equipment. However, these methods typically employ macroscopic hydrological models or static parameter assumptions, failing to adequately characterize key elements such as the complex micro-topographical structure within substations, including local slope variations, the distribution of impermeable underlying surfaces, and the spatial layout of drainage networks. This results in limited accuracy in predicting water depth and spread.

[0003] Current technologies generally neglect the dynamic regulatory role of initial soil deficit on the initial runoff generation process during rainfall, and fail to incorporate bottom elevation within the discharge outlet and seepage losses in the drainage pressure differential calculation. This results in significant discrepancies between simulated water receding conditions and actual conditions. Furthermore, existing emergency resource allocation largely relies on list-based matching, lacking spatial guidance and operational safety constraints that incorporate real-time water accumulation trends, hindering precise and efficient disaster relief decision-making. In addition, once deployed, monitoring models remain static and cannot be adjusted online based on actual water accumulation feedback, limiting the system's adaptability in complex and variable environments.

[0004] To address the aforementioned issues, there is an urgent need for a flood intelligent monitoring solution that can deeply integrate the microscopic physical constraints of substations, dynamically characterize hydrological processes, and possess spatial intelligent scheduling and model self-evolution capabilities, in order to achieve a shift from passive alarm to proactive risk prediction. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent monitoring method for flooding in substations, which can effectively solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] In a first aspect, a method for intelligent monitoring of flooding in substations is characterized by the following steps:

[0008] Construct a multi-dimensional influencing factor library, which includes geomorphic parameters, pipeline physical parameters, and environmental water consumption parameters for each functional zone of the substation. Among them, the pipeline physical parameters include at least the inner bottom elevation of each discharge outlet and the seepage loss rate of the pipeline.

[0009] Acquire rainfall forecast data and combine it with a multi-dimensional influencing factor library to perform dynamic water balance simulation. The simulation process includes: calculating the initial soil deficit water depth based on the number of consecutive rainless days in the previous period, and dynamically correcting the runoff coefficient at the beginning of rainfall based on the initial soil deficit water depth; at the same time, calculating the net drainage volume based on the effective pressure difference between the current water level and the bottom elevation of the discharge outlet, combined with the seepage loss rate of the pipeline.

[0010] Based on the net increase in water volume obtained from dynamic water balance simulation, combined with the high-precision three-dimensional digital model of the substation, three-dimensional water accumulation mapping is performed to generate a water level height field that evolves over time.

[0011] By comparing the water level field with the foundation elevation of key power equipment, equipment at risk of flooding is identified. Based on the current net increase in water volume, the total emergency drainage demand is calculated. A combination of drainage equipment that meets this total demand is matched from the emergency resource database. At the same time, using the centroid of water accumulation as the target point and combining it with the preset prohibited operation area, the layout coordinates of each drainage equipment are optimized, and a dispatch instruction containing the layout coordinates and the estimated arrival time is generated.

[0012] Preferably, the construction steps of the multidimensional impact factor library include:

[0013] The total area of ​​the station area was extracted by the geographic information system, the local slope ratio was obtained by gradient calculation through the digital elevation model, the proportion of impermeable underlying surface was obtained by fusing lidar point cloud and UAV oblique photography images and using support vector machine classification, and the soil hydraulic conductivity was obtained by field measurement or database call and the confidence level was marked.

[0014] The inner bottom elevation of each discharge outlet is extracted by total station measurement or building information model. The maximum depth of the pipeline is determined by three-dimensional pipeline model combined with on-site verification. The seepage loss rate of the pipeline is determined by referring to the specifications or water injection test based on the pipe material type and service life.

[0015] Vegetation cover type was retrieved by inverting multispectral images from drones, and the average daily water consumption per unit area was calculated by combining the reference crop evapotranspiration and crop coefficient provided by the weather station.

[0016] Preferably, in the dynamic water balance simulation, the step of dynamically correcting the runoff coefficient includes:

[0017] The runoff coefficient is set to a baseline value equal to the proportion of the impermeable underlying surface.

[0018] Determine whether the cumulative rainfall is less than the initial water deficit depth of the soil. If so, increase the runoff coefficient according to the attenuation function without exceeding the preset upper limit. If not, restore the runoff coefficient to the baseline value.

[0019] The initial water deficit depth of the soil is calculated based on the soil saturation water content and historical evaporation, or determined by empirical formula based on the number of consecutive rainless days in the preceding period.

[0020] Preferably, the calculation steps for the net drainage volume include:

[0021] Traverse all discharge outlets within the station area, select the discharge outlet with the lowest inner bottom elevation that is lower than the current water level as the effective drainage outlet, and calculate the difference between its inner bottom elevation and the current water level as the effective pressure difference;

[0022] When the effective pressure difference is greater than zero, the theoretical drainage capacity is calculated based on the orifice outflow principle. Then, the product of the seepage loss rate of the pipeline and the total length of the drainage pipeline is deducted to obtain the net drainage volume.

[0023] Preferably, the three-dimensional water accumulation mapping step includes:

[0024] The surface of the station area is divided into regular grids, and each grid records the ground elevation, slope and soil hydraulic conductivity;

[0025] The net increase in water volume at each time step is evenly distributed to each grid to obtain the initial water accumulation volume;

[0026] The D8 flow model is used to distribute water flow based on the ground slope and soil hydraulic conductivity between each grid cell. The process is iterated until a steady state is reached, and the final water accumulation volume of each grid cell is obtained.

[0027] Based on the water volume of each grid, the centroid coordinates of the water accumulation area are calculated, and the water spread boundary is determined by the flood filling algorithm, ultimately generating the water level height field.

[0028] Preferably, the steps for matching and optimizing the layout coordinates of the emergency drainage equipment combination include:

[0029] Available equipment is selected from the emergency resource pool and sorted by rated flow rate. A greedy algorithm is then used to select the minimum equipment combination that meets the required flow rate.

[0030] The station area ground is discretized into a grid. Using the coordinates of the water accumulation centroid as the target, the Euclidean distance from all feasible grid points to the centroid is calculated. After excluding grid points that fall into the prohibited operation area, the grid point with the smallest distance is selected as the layout coordinate.

[0031] The straight-line distance is calculated based on the coordinates of the equipment storage point and the deployment coordinates. The estimated arrival time is estimated by combining the average travel speed. At the same time, a path planning algorithm is used to generate a recommended travel path from the storage point to the deployment point.

[0032] Preferably, it also includes steps of online model calibration and evolution:

[0033] Real-time access to ultrasonic water level sensor data deployed in the station area; calculation of root mean square error between simulated water depth and measured water depth.

[0034] When the error exceeds the preset threshold, a manual interactive parameter correction slider is provided in the 3D visualization interface to adjust the local soil hydraulic conductivity or the global pipe and canal seepage loss rate.

[0035] Upon receiving the confirmation instruction, the current simulation is paused, the corrected parameters are reloaded, and the recalculation of the station's water accumulation evolution is initiated from the moment the rainfall begins, updating the water level field and emergency dispatch plan.

[0036] Preferably, the recalculation process is accelerated in parallel using a graphics processor:

[0037] The station area grid is divided into multiple sub-regions, each of which is assigned to an independent GPU thread block. The sub-region data is cached using shared memory for parallel computation, and the water accumulation evolution prediction for the next 2 hours is completed within 30 seconds.

[0038] Preferably, for different specific scenarios, at least one of the following adaptive adjustments is also included:

[0039] For old substations with missing pipeline data, ground-penetrating radar is used to detect the direction and depth of the pipelines, and pressure sensors are used to monitor and estimate the bottom elevation of the discharge outlet. At the same time, a siltation coefficient is introduced to correct the net drainage volume.

[0040] For coastal substations with high groundwater levels, groundwater level monitoring data is collected, groundwater backwater correction is introduced into the calculation of surface infiltration loss, and lateral seepage calculation based on Darcy's law is added in the process of determining the water spread boundary.

[0041] For scenarios with insufficient emergency resources, sandbags are added as emergency resources. The number of sandbags is automatically calculated based on the predicted water depth and the length of the area to be protected, and the shortest transportation route is planned.

[0042] Secondly, a substation flood intelligent monitoring system includes:

[0043] The multidimensional influencing factor library construction module is used to acquire and store the geomorphological parameters, pipeline physical parameters and environmental water consumption parameters of each functional zone of the substation. The pipeline physical parameters include at least the inner bottom elevation of each discharge outlet and the pipeline seepage loss rate.

[0044] The dynamic water balance extrapolation module is used to perform time-discrete water conservation calculations based on rainfall forecast data and the multi-dimensional influencing factor library. This module includes: a soil initial deficit water depth calculation unit, used to calculate the soil initial deficit water depth based on the number of consecutive rainless days in the previous period, and dynamically correct the runoff coefficient based on this value; and a net drainage volume calculation unit, used to calculate the net drainage volume based on the effective pressure difference between the current water level and the bottom elevation of the discharge outlet, combined with the seepage loss rate of the pipeline.

[0045] The three-dimensional water accumulation mapping module is used to generate a water level height field that evolves over time based on the net increase in water volume obtained through deduction and combined with a high-precision three-dimensional digital model of the substation.

[0046] The emergency resource scheduling module is used to determine the risk of equipment flooding based on the water level field and the foundation elevation of key power equipment, calculate the total emergency drainage demand based on the net increase in water volume, match drainage equipment combinations that meet the demand from the emergency resource database, and optimize equipment layout coordinates based on the water accumulation centroid as the target point and the prohibited operation area to generate scheduling instructions.

[0047] The online model calibration module is used to access sensor measured data in real time, receive parameter correction instructions through the human-computer interaction interface, and trigger the recalculation of the water accumulation evolution of the entire station to update the water level field and scheduling scheme.

[0048] In summary, this application includes at least one of the following beneficial technical effects:

[0049] 1. This invention constructs a multi-dimensional influencing factor library containing micro-physical parameters such as the bottom elevation of the discharge outlet and the seepage loss rate of the pipeline, and combines it with a dynamic correction mechanism based on the initial soil deficit based on the number of rainless days in the previous period. This enables accurate physical simulation of the water accumulation evolution process under the complex micro-topography of the substation, improves the accuracy of the prediction of initial runoff and water receding during rainfall, and avoids the false alarm problem caused by traditional static parameters.

[0050] 2. This invention calculates the net drainage volume based on the effective pressure difference between the real-time water level and the bottom marker of the discharge outlet, and optimizes the layout coordinates of drainage equipment by using the centroid of the water accumulation as a guide and combining the constraints of prohibited operation zones. This realizes the transformation of emergency resources from list-based matching to precise scheduling at the spatial coordinate level, improving the efficiency of emergency rescue while ensuring operational safety, and providing a scientific basis for emergency decision-making.

[0051] 3. This invention achieves online parameter correction through a human-computer interactive slider and utilizes GPU parallel acceleration technology to quickly recalculate the water accumulation of the entire station. At the same time, it employs a machine learning model to automatically evolve the initial values ​​of soil parameters based on historical event data, realizing a closed-loop evolution of the monitoring model from static operation to dynamic adaptation, effectively improving the system's generalization ability and long-term reliability in different substation environments. Attached Figure Description

[0052] Figure 1 This is a schematic diagram of the overall technical solution architecture of the intelligent flood monitoring method for substations proposed in this invention;

[0053] Figure 2 This is a schematic diagram of the core principle framework of the coupling between dynamic water balance deduction and three-dimensional spatial water accumulation mapping in this invention;

[0054] Figure 3 This is a flowchart of the automatic matching and scheduling logic for emergency resources in this invention;

[0055] Figure 4 This is a schematic diagram of the multi-level interaction relationship and data flow of the online model correction and evolution mechanism in this invention. Detailed Implementation

[0056] To further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the following detailed description of specific embodiments based on the present invention is provided in conjunction with the accompanying drawings and preferred embodiments.

[0057] Example 1

[0058] A method for intelligent monitoring of flooding in substations includes the following steps:

[0059] S1 constructs a multi-dimensional influencing factor library. Its execution process includes a geomorphological parameter collection sub-process, a pipeline physical parameter integration sub-process, and an environmental water consumption parameter input sub-process. The three processes are aligned and fused using a unified spatiotemporal coordinate system.

[0060] S101, collect geomorphic parameters to obtain key geomorphic features that affect surface runoff and infiltration processes within the substation, specifically including four parameters: total area of ​​the station area, local slope ratio, proportion of impermeable underlying surface, and soil hydraulic conductivity.

[0061] The total area of ​​the station area is measured in square meters. The boundary vector map of the high-precision geographic information system (GIS) is extracted, and the error is controlled within ±0.5%.

[0062] The local slope ratio is expressed as a percentage and is obtained by calculating the gradient of the digital elevation model (DEM) using the Sobel operator, with a spatial resolution of 0.1 m × 0.1 m grid cells.

[0063] The proportion of impermeable underlying surfaces was obtained through a fusion model of LiDAR point cloud and UAV oblique photogrammetry images, with a point cloud density of no less than 200 points / square meter and a ground sampling distance (GSD) of better than 2 cm. Subsequently, a support vector machine (SVM) surface material classification algorithm was used to perform pixel-level identification of four typical underlying surfaces: asphalt, concrete, lawn, and bare soil. The classification accuracy reached over 95% after cross-validation. Finally, the coverage area ratio of each type was calculated, and the numerical value of the impermeable underlying surface proportion was output.

[0064] Soil hydraulic conductivity is measured in millimeters per hour and is obtained from laboratory measurements of soil borehole samples taken on-site or through interpolation from historical geological survey reports. If no measured data is available, the default value in the regional soil type database is used, such as the China Soil Database or the World Soil Database. The confidence level is marked in subsequent use based on the reliability of the data source. For example, if the confidence level of the measured data is 1, the confidence level of the database default value is 0.7.

[0065] Soil saturated water content θ sat Soil texture type is determined based on on-site soil sampling and testing. Soil texture is classified into categories such as sandy soil, loam, and clay by sieve analysis or hydrometer method. Then, the corresponding typical saturated water content (THS) value is obtained by referring to commonly used soil physics handbooks or databases, such as the USDA Soil Texture Classification Table, in millimeters. If soil sampling is unavailable, the texture classification information can be obtained from a regional soil type database using the same method. In subsequent use, the confidence level should be labeled according to the data source.

[0066] S102 integrates the physical parameters of the pipeline network, targeting the key physical attributes of the drainage network inside the substation, including the bottom elevation of each discharge outlet, the maximum depth of the pipeline, and the seepage loss rate.

[0067] The bottom elevation of each discharge outlet is measured in meters and is relative to the 1985 National Elevation Datum. It is measured by total station or extracted from the building information model (BIM) as-built model, with an accuracy better than ±2 millimeters.

[0068] The maximum depth of the pipeline refers to the vertical distance from the ground surface to the bottom of the pipeline. This value is determined by combining the three-dimensional pipeline model with the on-site verification data.

[0069] Permeation loss rate (see page) rate Defined as the leakage volume per unit length of pipe per unit time, in cubic meters per meter per hour. This value can be set according to the pipe material type, such as concrete, HDPE, cast iron, and service life. Specific values ​​can be found in the empirical values ​​for pipe leakage given in the "Outdoor Drainage Design Code," or determined through on-site water injection tests. For example, 0.0005 cubic meters per meter per hour can be used for newly laid HDPE pipes, and 0.003 cubic meters per meter per hour for concrete pipes that have been in service for more than 15 years. Other cases can be determined by linear interpolation based on pipe material and service life, or by consulting relevant manuals. This parameter is stored in the pipe network attribute table and is bound to a unique identifier ID for each pipe section.

[0070] Effective flow area A of each discharge port orificeThe effective flow area is calculated based on the geometry and dimensions of the discharge outlet. For circular pipes, the effective flow area equals π multiplied by one-quarter of the square of the pipe's inner diameter; for rectangular or square discharge outlets, the effective flow area equals the width multiplied by the height. The geometric dimensions of the discharge outlet can be obtained by consulting design drawings or through on-site measurement, with a measurement accuracy better than ±5 mm. This parameter is stored in the pipe network attribute table and is bound to the unique identifier ID of the corresponding discharge outlet.

[0071] The total length L of the drainage pipes pipe L is the sum of the lengths of all pipes and channels within the station area that may be involved in drainage. The length of each pipe and channel can be measured from the 3D pipeline model or as-built drawings, or obtained through on-site measurement. During parameter initialization, all pipe and channel records in the pipeline attribute table are traversed, and their length fields are summed to obtain L. pipe The value, in meters, is stored as a global parameter in the configuration of the water balance simulation engine.

[0072] S103, Enter environmental water consumption parameters, mainly considering the impact of vegetation cover on water consumption in the early stages of rainfall, including vegetation cover type and its average daily water consumption per unit area.

[0073] Vegetation cover type is determined by inverting the normalized vegetation index (NDVI) from UAV multispectral imagery. The NDVI value ranges from -1 to 1. Based on empirical thresholds, it is divided into three categories: trees, shrubs, and lawns. Generally, areas with an NDVI greater than 0.6 are identified as trees, those between 0.3 and 0.6 are identified as shrubs, and those less than 0.3 are identified as lawns.

[0074] The average daily water consumption per unit area is dynamically calculated based on the reference crop evapotranspiration ET0 and crop coefficient Kc provided by the local meteorological station. ET0 is in millimeters per day and can be obtained directly from the meteorological station or calculated using the Penman formula based on meteorological factors such as temperature, humidity, and wind speed. Kc is taken as 0.7 for lawns, 0.9 for shrubs, and 1.1 for trees. The formula for calculating the average daily water consumption is Kc × ET0, in millimeters per day. This value is automatically updated and written to the environmental parameter cache every morning to ensure that the water consumption parameter can reflect recent changes in meteorological conditions.

[0075] S104 After completing the collection and integration of the above three types of parameters, organize the parameters into a structured data table according to the actual functional zoning of the substation, such as the main transformer area, the power distribution equipment area, and the cable interlayer area.

[0076] Each partition corresponds to a JSON-formatted parameter object, containing fields such as partition identifier (area_id), total area (total_area), impervious ratio (impervious_ratio), average slope (slope_mean), soil hydraulic conductivity (hydraulic_conductivity), list of bottom elevations inside discharge outlets (invert_elevation_list), maximum pipe depth (max_pipe_depth), seepage rate (seepage_rate), vegetation type (vegetation_type), and daily evapotranspiration (daily_evapotranspiration).

[0077] The aforementioned parameter objects are pushed to the water balance inference engine via message queues such as RabbitMQ, thereby completing the construction of the multi-dimensional influencing factor library.

[0078] In addition, soil saturated water content θ sat The root, as a global parameter, is pushed to the water balance simulation engine along with the partition data; the total length L of the pipes and channels involved in drainage... pipe The system automatically calculates and summarizes the data after loading the pipeline attribute table, without needing to pass it through a partition JSON object.

[0079] Through the above steps, the multidimensional impact factor database is constructed. This database consists of two parts: first, JSON parameter objects organized by functional zones, containing the topography, basic pipe network attributes, and environmental water consumption parameters for each zone; second, global parameters, including soil saturation moisture content and the total length of participating drainage pipes and canals automatically summarized from the pipe network attribute table. The pipe network attribute table is an independent table stored in the database, recording the detailed physical parameters of each pipe segment, and is one of the data sources for the multidimensional impact factor database.

[0080] S2 executes dynamic water balance simulation, establishes time-discrete water conservation equations, and introduces a dual dynamic correction mechanism.

[0081] S201: Acquire rainfall forecast data and set time parameters; connect in real-time to the rainfall forecast for the future period issued by the meteorological department; divide the forecast duration into several consecutive time steps, with the step size Δt of each time step fixed at 10 minutes; rainfall intensity P(t)... i Input in millimeters every 10 minutes, t i This represents the start time of the i-th time step.

[0082] S202, Initialize the basic parameters required for dynamic water balance simulation. Specifically:

[0083] The total area A of the station area, in square meters, is retrieved from the multidimensional influencing factor database, and the soil saturated moisture content θ is also retrieved. satThis parameter is obtained by referring to the soil texture type determined by on-site soil sampling, and then looking up the corresponding typical value of saturated water content in millimeters from commonly used soil physics manuals or databases such as the USDA Soil Texture Classification Table.

[0084] Read the sum of the average daily evaporation over the past 30 days, E hist The data, provided by the local weather station, is in millimeters and represents the proportion of impervious subsurface surfaces. ratio This is used to set the initial value φ0 of the dynamic runoff coefficient. The setting rule is that φ0 is equal to the proportion of the impermeable underlying surface. For example, when the impermeable proportion is 25%, φ0 is 0.25.

[0085] Read the bottom elevation H of each discharge port invert The unit is meters, and the effective flow area is A. orifice The unit is square meters, and the total length L of the drainage pipes and channels involved is... pipe The unit is meters, and the pipeline seepage loss rate is shown on page [page number]. rate The unit is cubic meters per meter per hour. The soil hydraulic conductivity K is read in millimeters per hour, and the current water level H is set. water The initial value is 0.

[0086] S203 calculates the initial soil water deficit depth I0, used to correct for infiltration capacity at the beginning of rainfall. This is based on the number of consecutive rainless days D prior to rainfall. dry If the number of days is greater than or equal to 5, then I0 calculation is initiated, and the calculation formula is as follows:

[0087] I0=min(θ sat ×0.6,E hist ×0.8);

[0088] Where I0 is in millimeters, and the coefficients 0.6 and 0.8 are empirical values, representing the upper limit of the proportion of pore water that can be released after the soil is saturated and the efficiency factor of the conversion of the previous evaporation into the soil's water storage capacity, respectively. This value indicates the upper limit of the amount of additional water that the soil can absorb before the start of rainfall.

[0089] S204, for each time step t i The dynamic runoff coefficient φ(t) is executed sequentially. i () Correction.

[0090] Firstly, the baseline value φ0 of the dynamic runoff coefficient is directly determined by the proportion of the impermeable underlying surface, that is, φ0 equals the proportion of the impermeable underlying surface. Then, the current cumulative rainfall... The relationship between the sum from the first time step to the current step and I0: If the value is less than I0, it indicates that the soil is not yet saturated. At this point, the runoff coefficient φ(t) is... i The adjustment is made dynamically according to the attenuation function, that is:

[0091] ;

[0092] This calculation ensures φ(t) i The index gradually increases with increasing rainfall, and its value does not exceed a preset upper limit of 0.1 before the soil becomes saturated. Here, the index of 1.2 and the upper limit of 0.1 are empirical parameters obtained by fitting typical hydrological experimental data. They can be calibrated and adjusted based on historical water accumulation records in the station area during practical applications. If it is greater than or equal to I0, then φ(t) i It is restored to φ0.

[0093] S205, Calculate the real-time drainage capacity Q(t) i This capacity depends on the effective pressure difference ΔH between the current water level and the bottom elevation of the discharge outlet, i.e.:

[0094] ΔH=H water (t i )-H invert ;

[0095] Where H water (t i The water level at the start of the current time step is expressed in meters (H). invert The inner bottom elevation of the selected discharge outlet, also in meters.

[0096] At each time step, all pre-collected discharge outlets within the station area are traversed, and the discharge outlet with the lowest inner bottom elevation below the current water level is selected as the valid discharge outlet. The inner bottom elevation of this discharge outlet is H. invert Its effective flow area is denoted as A. orifice The unit is square meters.

[0097] If the elevation of the inner bottom without an outlet is lower than the current water level, then the drainage capacity Q(t) i A forced zero value indicates that the accumulated water cannot be drained by gravity. When ΔH is greater than zero, the drainage capacity is calculated using the orifice outflow formula, specifically:

[0098] ;

[0099] In the formula C d The flow coefficient is set to 0.62, and g is the acceleration due to gravity, set to 9.81 m / s². The calculated Q(t) i The unit is cubic meters per hour.

[0100] Calculated Q(t) i The actual net drainage volume Q used to reduce water accumulation needs to be deducted from the seepage losses of the pipe network itself. net (t i ).

[0101] Permeation loss rate (see page) rate Defined as the leakage volume per unit length per unit time of a pipe, in cubic meters per meter per hour. This value can be determined by referring to the empirical value in the "Outdoor Drainage Design Code" or by on-site water injection test, depending on the pipe type and service life. For example, 0.0005 cubic meters per meter per hour can be used for newly laid HDPE pipes, and 0.003 cubic meters per meter per hour can be used for concrete pipes that have been in service for more than 15 years.

[0102] The formula for calculating net drainage volume is:

[0103] Q net (t i )=Q(t i -seepage rate ×L pipe ;

[0104] Where L pipe The total length of the drainage pipes and channels involved is expressed in meters. The result is Q. net (t i The unit is still cubic meters per hour.

[0105] S206, Calculate the surface infiltration loss I(t) i (hereinafter referred to as infiltration loss) refers to the amount of water that infiltrates into the soil through the surface during rainfall, which is constrained by both the soil's hydraulic conductivity and its current water storage capacity.

[0106] Soil hydraulic conductivity K is expressed in millimeters per hour, the total area of ​​the station area A is expressed in square meters, and the time step Δt is fixed at 10 minutes, or 1 / 6 of an hour. Within each time step, the maximum possible infiltration is determined by the soil hydraulic conductivity, calculated using the following formula:

[0107] ;

[0108] Multiplying by 0.001 converts millimeters to meters, making I max The unit is cubic meters.

[0109] The initial soil water deficit depth I0, in millimeters, is calculated from S203, and the initial soil water storage volume V is... soil_remain =I0×A×0.001, in cubic meters. At each time step, the actual infiltration rate I(t) i ) takes I max Compared with the current remaining soil water storage capacity V soil_remain The smaller value in, that is:

[0110] ;

[0111] Then update the remaining soil water storage capacity:

[0112] ;

[0113] This process ensures that once the soil's water storage capacity is depleted, the infiltration rate in subsequent time steps is limited only by the hydraulic conductivity.

[0114] S207, Calculate the net increase in water volume V(t) at the current time step. i Net increase in water volume refers to the amount of water added to the station area due to rainfall during that time step. The calculation formula is:

[0115] ;

[0116] Wherein, P(t) i ) represents the rainfall within this time step, in millimeters, directly read from rainfall forecast data; A represents the total area of ​​the station area, in square meters; φ(t) i I(t) represents the dynamic runoff coefficient at the current time step, which is dimensionless; i Q represents infiltration loss, in cubic meters; net (t i ) represents the net drainage volume in cubic meters per hour; Δt represents the time step in hours, which is 1 / 6 of an hour here.

[0117] In the above formula, P(t) i )×A×φ(t i )×0.001 converts rainfall into runoff volume in cubic meters.

[0118] Calculation result V(t) i A positive value indicates an increase in water accumulation; a negative value indicates a decrease in water accumulation.

[0119] S208, Update the total volume of accumulated water V total Let the total volume of water at the end of the previous time step be V. total (t i If the total volume of water at the end of the current time step is:

[0120] ;

[0121] Initial time V total (0)=0.

[0122] The total volume of accumulated water will be used as input for the subsequent three-dimensional spatial mapping step, and will be combined with the topographic distribution of the station area to transform it into a new water level distribution H(x,y,t). i+1 This water level distribution will serve as a reference when calculating drainage capacity in the next time step.

[0123] After all time steps have been iterated, the complete net increase in water volume sequence V(t) is output.i It is used for three-dimensional space mapping.

[0124] S3 performs three-dimensional spatial water accumulation mapping, which is achieved by coupling a high-precision three-dimensional digital model with a hydraulic diffusion algorithm.

[0125] S301 constructs a high-precision 3D digital model of the substation. It uses Bentley ContextCapture software to generate a real-scene mesh model based on UAV oblique photogrammetric images. The model has more than 50 million vertices and a texture resolution of 4K.

[0126] During the model building process, the basic elevation attributes of key power equipment, including 10kV switchgear, main transformer, GIS combined electrical equipment, etc., are embedded simultaneously. Their basic elevations are stored in the form of independent point cloud tags, and the elevation accuracy is controlled within ±5 mm, so as to provide a fine geometric basis for the subsequent spatial distribution of water accumulation.

[0127] S302, the surface is divided into regular grids and attributes are recorded. The entire station area is divided into regular grid units of 0.5 meters by 0.5 meters. Each grid records three core physical attributes: ground elevation Z(x,y) is extracted from the 3D model, ground slope S(x,y) is calculated from the elevation data using the Sobel operator, and soil hydraulic conductivity K(x,y) is obtained by interpolation of the soil hydraulic conductivity of the corresponding partition in the multidimensional influence factor library. All grid attributes are stored in a two-dimensional matrix for easy access.

[0128] S303, allocate the net increase in water volume to the grid cells. At the beginning of each time step, use the net increase in water volume V(t) calculated in S2 for that time period. i The water volume is evenly distributed across all surface grid cells, meaning each grid cell receives an initial increase in water volume.

[0129] ;

[0130] Where N is the total number of grid cells in the station area.

[0131] Add this increment to the water accumulation volume of each grid cell at the previous time step to obtain the water accumulation volume V of each grid cell at the current time step. i The uniform distribution assumption assumes that runoff is spatially uniform, which aligns with the actual situation of uniform rainfall in small-scale regions.

[0132] S304 performs inter-grid water overflow distribution based on the D8 flow direction model. For each grid cell with water accumulation, its eight neighborhood directions are first evaluated. For each neighborhood direction, the directional slope D from the current grid cell to that neighboring grid cell is calculated. dir The formula is:

[0133] ;

[0134] Among them, Z center Z represents the ground elevation of the current raster. neighbor d represents the ground elevation of the neighboring grid, and d is the distance between the center points of adjacent grids, which is 0.5 meters.

[0135] If D dir If the value is less than or equal to 0, it means that a gravitational outflow cannot be formed in that direction, so this direction is ignored. From all D... dir Among the directions with a slope greater than 0, select the direction with the largest slope and record these directions as the candidate direction set.

[0136] If there is only one candidate direction, the entire water volume of the current grid is transferred to the neighboring grid in that direction at once. After the transfer, the water volume of the current grid is set to zero, and the water volume of the target grid is increased by the corresponding amount of water.

[0137] If there are multiple candidate directions with equal slopes, then the directions are weighted and allocated according to their hydraulic conductivity.

[0138] The proportion of water volume allocated to each direction is determined by weight W. dir The weighting formula is as follows:

[0139] ;

[0140] Among them, D dir K represents the directional slope in that direction. dir The soil hydraulic conductivity of this neighborhood grid is given by the denominator, which is the sum of the products of the slope and hydraulic conductivity of all candidate directions.

[0141] Since the slopes of the candidate directions are equal, the weights can be simplified to K. dir / ∑K k According to this weight, the water volume of the current grid is divided and transferred to the neighboring grids of each candidate direction. After the transfer, the water volume of the current grid is set to zero, and each target grid receives the corresponding water volume and accumulates it.

[0142] After completing one transfer, the above process is repeatedly executed iteratively on all water accumulation grids in the entire station until all water accumulation grids no longer change, that is, a stable state is reached.

[0143] S305, calculate the centroid location of the water accumulation area. After completing the overflow distribution, calculate the final water accumulation volume V of each grid. i The formula for calculating the coordinates of the centroid of the accumulated water is:

[0144] ;

[0145] Where, x i and y iV represents the coordinates of the grid center point. i Given the water volume of a grid, summation is performed for all grid cells with a water volume greater than zero.

[0146] If the water volume of all grid cells is zero, there is no water accumulation area, and there is no need to calculate the centroid. The centroid coordinates reflect the spatial center location of the total water volume and are used for subsequent optimization of emergency resource deployment.

[0147] S306, determine the boundary of water accumulation spread.

[0148] First, determine the starting grid. Since the centroid coordinates may not fall exactly at the center of a grid, the grid containing the centroid coordinates is taken as the starting grid. If the water volume of the grid is zero, the grid closest to the centroid is selected from all water-filled grids as the starting grid.

[0149] Starting from the initial grid, the flood fill algorithm is used to recursively expand the water accumulation area. For the current water accumulation grid, the water surface elevation H water The ground elevation Z(x,y) of the grid and the water depth d i Adding them together, we get d. i =V i / S grid S grid The area of ​​a single grid cell is 0.25 square meters.

[0150] Then check the eight neighboring rasters of the current raster, and for each neighboring raster, obtain its ground elevation Z. neighbor If Z neighbor The elevation is lower than the current water level minus the safety margin δ, which satisfies H. water ≥Z neighbor If the condition is +δ, then the neighboring grid cell is determined to be flooded, and it is added to the flooded area and recursively expanded as a new current grid cell; if the condition is not met, then expansion along that direction is stopped.

[0151] The safety margin δ is usually taken as 0.05 meters to account for water surface fluctuations and model errors.

[0152] The expansion process continues until all potentially flooded grids have been traversed. At this point, all grids marked as flooded constitute a waterlogged area, the boundary of which is determined by the intersection of flooded and unflooded grids.

[0153] S307, generate water level height field and verify accuracy.

[0154] The above-mentioned overflow allocation and boundary determination process is executed once every 10 minutes, consistent with the time step of the water balance simulation, thereby generating a water level height field H(x,y,t) that evolves over time, which is the water surface elevation value of each grid.

[0155] To verify the model's accuracy, ultrasonic water level sensors were deployed at key locations within the substation to measure the actual water depth H in real time. obs At each sensor installation point, the simulated water depth H of the corresponding grid is extracted. sim This allows for the calculation of the absolute error between the simulated water depth and the measured water depth. The simulated water depth H is... sim The formula is as follows:

[0156] ;

[0157] The average absolute error of all measuring points should be controlled within ±5 cm, and the maximum error should not exceed ±15 cm to meet the requirements of engineering applications. If the error exceeds the limit, the model correction mechanism in step S5 will be triggered.

[0158] S4 triggers the automatic matching mechanism for emergency resources. The emergency response resource database involved in this step is a dedicated database independent of the multidimensional impact factor database. It is used to store detailed information on emergency resources such as drainage equipment, sandbags, and gates, including fields such as equipment number, type, rated flow, storage point coordinates, and prohibited operation areas. This resource database, together with the multidimensional impact factor database, supports the complete functionality of this system.

[0159] S401, Perform equipment flooding risk assessment. First, obtain a list of critical power equipment from the equipment ledger or 3D model. Each piece of equipment includes a unique identifier, name, spatial coordinates (x and y), and foundation elevation (H). base The basic elevation usually refers to the height of the equipment base or the lowest electrical component relative to the 1985 National Elevation Datum, in meters.

[0160] Then the water level height field H(x,y,t) generated in step S3 is regarded as regular raster data, and each raster records the water surface elevation at that location.

[0161] For each device, locate the corresponding grid based on its coordinates, and read the water surface elevation value H(x,y,t) of that grid. If the water surface elevation satisfies H(x,y,t)≥H base -δ safety , where δ safety To set a safety margin, typically 0.1 meters, the equipment is considered to be at risk of being submerged.

[0162] Once the assessment is complete, the system will automatically highlight all flooded equipment in the 3D visualization interface and mark each flooded device with a red warning icon to alert maintenance personnel to pay close attention.

[0163] S402, calculate the total emergency drainage demand based on the net increase in water volume V(t) at the current time step. i Given the given time step Δt, calculate the required total drainage flow rate Q. requiredNet increase in water volume V (t) i The unit is cubic meters, calculated in step S207, and may be positive or negative.

[0164] If V(t) i If Q ≤ 0, it means that the water level has not increased or has even decreased during the current period, and there is no need to activate emergency drainage. required =0.

[0165] If V(t) i If ) > 0, then the required flow rate is calculated using the following formula:

[0166] ;

[0167] Where Δt is the time step, which is fixed at 10 minutes here and needs to be converted to hours, i.e., Δt = 10 / 60 = 1 / 6 hour. α is the safety margin coefficient, which is fixed at 0.2, representing an additional 20% redundancy based on the theoretical drainage flow.

[0168] Calculation result Q required The unit is cubic meters per hour, used to guide the selection and combination of emergency resources.

[0169] S403, Search the emergency resource database and match the minimum equipment combination.

[0170] The emergency response resource database is stored using a relational database. Each resource record represents an available drainage device and includes the following fields: equipment_id (unique identifier), type (e.g., drainage pump truck or mobile submersible pump), rated_flow (in cubic meters per hour), power (in kilowatts), deployment_radius (in meters), forbidden_zones (a set of polygon coordinates), and storage point coordinates (depot_x and depot_y in meters).

[0171] When matching devices, firstly, all devices in the resource pool that are available and whose deployment radius can cover the site area or at least reach the site area are selected as candidates. Then, based on Q... required Retrieve from candidate devices the sum of the rated flow rates of all selected devices, ∑rated. flow ≥Q required The equipment combination should prioritize the fewest possible equipment while ensuring that the rated flow rate of each individual device is as close as possible to Q. required The combination of .

[0172] The specific retrieval logic employs a greedy algorithm: candidate devices are sorted from largest to smallest rated traffic, and the device with the largest traffic is selected sequentially. If its rated traffic already meets the requirement, that single device is directly selected; if a single device cannot meet the requirement, the next largest device is added, until the accumulated traffic meets Q. required The final output is a list of device combinations along with their respective models and quantities. This strategy can, in most cases, yield combinations with a smaller number of devices, reducing the complexity of on-site deployment.

[0173] If the sum of the rated flow rates of all available devices is still less than Q required If the output is not found, all available devices will be displayed and a message indicating insufficient resources and the need for external support will be provided.

[0174] S404, Optimize equipment layout location, based on the centroid coordinates of the accumulated water x c y c For the target point, search for the optimal deployment location x within the station area. deploy y deploy The search space is limited to the ground of the station area, and the deployment points cannot be located in any prohibited work areas. These prohibited work areas include cable trenches, operating passages, and a 2-meter radius around fire hydrants. The polygon coordinates of these areas have been pre-stored in the forbidden_zones field of the emergency resource library.

[0175] To simplify the search, the station area ground is discretized into a regular grid of 0.5 meters by 0.5 meters, the same as in step S302. Each grid point represents a candidate deployment location. All grid points are traversed, and those located within the prohibited operation area are excluded. The Euclidean distance from each of the remaining feasible grid points to the centroid of the accumulated water is calculated, i.e., the distance is equal to the square root of x. i Subtract x c squared plus y i Subtract y c The square of.

[0176] Select the grid point with the smallest distance as the final deployment location x. deploy y deploy If all grid points are prohibited, select the non-prohibited grid point closest to the centroid and notify the maintenance personnel that the deployment location may be restricted.

[0177] S405, Generate and push scheduling instructions, integrating the device combination matched in step S403, the deployment coordinates determined in step S404, and the estimated arrival time of each device into scheduling instructions.

[0178] Calculating the estimated arrival time requires the coordinates of the storage location of each device, which are recorded in the emergency resource database. The corresponding field for each device is "depot". x and depot yThe unit is meters, consistent with the station area coordinate system. For each piece of equipment in the assembly, it is based on the coordinates x of its storage point. depot y depot With the coordinates x of the deployment point deploy y deploy Calculate the straight-line distance:

[0179] ;

[0180] Where d is in meters. Dividing the distance by the average speed, taken as an empirical value of 30 kilometers per hour (v = 30,000 meters per hour), yields the estimated arrival time in hours.

[0181] ;

[0182] If the storage point coordinates are not recorded in the resource library, the deployment radius field deployment_radius built into the device is used as the estimated value of distance d. If the deployment radius is also missing, 30 minutes is taken by default.

[0183] The dispatching instructions include the equipment model, quantity, specific deployment coordinates, estimated arrival time, and operational precautions. At the same time, a recommended travel path from the equipment storage point to the deployment point is generated. The A* algorithm is used to avoid buildings and prohibited work areas on the station area plan, plan the shortest feasible path, and overlay the equipment icon and the planned path in the 3D model.

[0184] In addition, dispatch instructions are pushed to the operation and maintenance dispatch platform through the standard API interface for on-duty personnel to dispatch and execute.

[0185] S5 enables online model correction and evolution, and its interaction mechanism is integrated into the 3D visualization platform.

[0186] The S501 receives real-time on-site monitoring data. The 3D visualization platform continuously receives two real-time data streams: one is a high-definition video surveillance stream deployed in the station area, with a video resolution of 1920 x 1080 pixels and a frame rate of 25 frames per second; the other is data from ultrasonic water level sensors installed in key water accumulation areas, with a sensor measurement accuracy of ±1 mm. These two data streams provide a true reference for the water accumulation status for subsequent deviation analysis.

[0187] S502 performs simulation and measurement deviation analysis, automatically calculating the root mean square error (RMSE) between the simulated and measured water depths at all ultrasonic water level sensor measuring points every 5 minutes.

[0188] For the j-th sensor, the measured water depth is denoted as H. obsj The grid simulation water depth H corresponding to its coordinates simjThe water level height field H(x,y,t) generated in step S307 is subtracted from the ground elevation Z(x,y) of the grid, i.e.:

[0189] ;

[0190] Where x j y j Let be the planar coordinates of the j-th sensor.

[0191] Suppose there are M sensors in total, then the formula for calculating the root mean square error is:

[0192] ;

[0193] If the calculated RMSE exceeds a preset threshold, such as 5 cm, it is determined that there is a significant deviation in the current model prediction. A semi-transparent deviation heatmap is immediately superimposed on the surface of the water accumulation area of ​​the 3D model. The redder the area in the heatmap, the greater the deviation, thus intuitively prompting maintenance personnel to pay close attention.

[0194] The S503 provides a manual interactive parameter calibration interface. Maintenance personnel can click on any water accumulation grid in the 3D scene to bring up the parameter calibration panel, which contains two adjustable sliders.

[0195] The first slider is used to adjust the soil hydraulic conductivity of the currently clicked grid. The initial value displayed by the slider is the current soil hydraulic conductivity value K(x,y) of the grid, which ranges from 1 to 50 mm / h with a step size of 0.1 mm / h. Maintenance personnel can drag the slider to modify the soil hydraulic conductivity of the grid. After modification, only the K value of the grid is updated, while the values ​​of other grids remain unchanged.

[0196] The second slider is used to adjust the seepage loss rate of the entire station area's pipeline network. The initial value displayed by the slider is the current global seepage loss rate (see page). rate The value ranges from 0.0001 to 0.005 cubic meters per meter per hour, with a step size of 0.0001 cubic meters per meter per hour. After the maintenance personnel drag the slider to modify this value, the seepage loss rate of all pipe sections will be updated to the new value.

[0197] Maintenance personnel can drag the slider to adjust the parameter values ​​based on on-site experience or temporary observation data. After the adjustment is completed, clicking the confirmation button will save the modified parameters and trigger the recalculation process in step S504.

[0198] S504, Perform parameter correction and recalculate the total station water volume. After the maintenance personnel drag the slider to adjust the parameters in the correction panel in step S503 and click confirm, first record the time step of the current water balance simulation, save the time step and all previous calculation results, and pause the calculation of subsequent time steps; then reload the corrected parameter values ​​into the water balance equation, specifically including updating the soil hydraulic conductivity of a single grid cell modified in step S503 to the attribute table of that grid cell, and updating the modified global pipe network seepage loss rate to the attribute table of all pipe segments.

[0199] Next, the recalculation of the water accumulation evolution of the entire station is initiated. The recalculation re-executes steps S2 and S3 from the moment the rainfall begins until the current paused time step, thereby obtaining the new water accumulation evolution result after correcting the parameters.

[0200] To accelerate the recomputation process, a parallel computing strategy based on the graphics processing unit (GPU) is adopted. The entire station grid is divided into several sub-regions according to spatial location. Each sub-region contains a continuous range of rows and columns. The number of sub-regions is dynamically determined according to the number of GPU computing cores. For example, for a typical GPU with 2560 CUDA cores, the station grid can be divided into 16×16 sub-regions. The size of each sub-region is about 1 / 16 of the original grid's row and column count. The computing task of each sub-region is assigned to a GPU thread block. Multiple threads within the GPU thread block jointly process the grid within that sub-region.

[0201] During the calculation process, each GPU thread block first loads the elevation data and soil hydraulic conductivity data of the corresponding sub-region from the video memory into the shared memory to reduce the latency of subsequent repeated access to video memory. Then, the threads within the GPU thread block calculate the water transfer of each grid in the sub-region in parallel according to the D8 flow model. The boundary data between adjacent sub-regions are synchronously exchanged through global memory.

[0202] After the recalculation is completed, the new water accumulation evolution results are compared with the original results, the water level height field and warning level in the three-dimensional interface are updated, and the emergency dispatch plan is generated again by calling step S4 according to the new water accumulation situation. The updated plan is then pushed to the operation and maintenance dispatch platform.

[0203] It should be noted that the entire recalculation process must ensure that the simulation of the water accumulation over the next two hours is completed within 30 seconds, which can be achieved by adjusting the sub-region size and GPU thread configuration.

[0204] S505 archives historical flood event data and automatically monitors rainfall processes. When there is no effective rainfall for two consecutive hours and the water depth measured by all sensors in the station area drops to below 5 centimeters, the rainfall event is considered to have ended.

[0205] After the event concludes, a complete record of the event will be archived in the historical flood event database. The archived content includes:

[0206] The time-series data of water depth measured by all ultrasonic water level sensors every 10 minutes during the event are used to form a time-depth curve.

[0207] The start and end times of an event, i.e., the duration from the first effective rainfall detected to the end of the event;

[0208] The soil hydraulic conductivity adjustment value and the pipeline seepage loss rate adjustment value and their corresponding grid positions are entered by the maintenance personnel in step S503;

[0209] The final deployment effect of the emergency equipment, i.e. whether the equipment flooding was successfully avoided, is recorded as "successfully avoided" or "flooding occurred". If flooding occurred, the list of flooded equipment and the flooding depth are recorded.

[0210] The archived data mentioned above will provide samples for subsequent model training.

[0211] S506 trains a predictive evolutionary model and updates default parameters by drawing all archived event records from the historical flood event database weekly as training samples.

[0212] For each event, construct a set of input features and corresponding output labels, where the input features include:

[0213] The number of consecutive rainless days in the pre-event period, i.e. the number of days from the end of the last rainfall before the start of the event to the start of the event;

[0214] The rainfall intensity for this event is calculated by dividing the total rainfall during the event by the duration of the event, and the unit is millimeters per hour.

[0215] The proportion of impermeable underlying surface in the station area is a fixed parameter, read from the multidimensional influencing factor library.

[0216] The output labels represent the parameter corrections that need to be optimized, including corrections for the initial soil water deficit depth I0 and the soil hydraulic conductivity K. The methods for determining these corrections are as follows:

[0217] For each historical event, the actual rainfall data of the event is input into the uncorrected original model to obtain a set of simulated water accumulation time series. Then, with the measured water accumulation time series as the objective, the Bayesian optimization algorithm is used to search for the adjustment values ​​of I0 and K that minimize the root mean square error between the simulation and the measurement. This adjustment value is the optimal correction amount corresponding to the event.

[0218] The input features of all events and their corresponding optimal corrections are combined into a training set to train the XGBoost regression model. The hyperparameters of the XGBoost regression model are set as follows: maximum tree depth is set to 6, learning rate is set to 0.1, number of subtrees is set to 100, and the remaining parameters use default values.

[0219] After training, the XGBoost regression model can predict recommended I0 and K corrections based on the number of drought days, rainfall intensity, and impermeability ratio in the lead-up to a new event. These predicted corrections are then automatically applied to the default parameter values ​​before the next rainfall event begins. In other words, the default values ​​of I0 and soil hydraulic conductivity are updated to the original default values ​​plus the predicted corrections, thereby enabling the continuous evolution of the model parameters.

[0220] Example 2

[0221] For the special application scenario of severe data gaps in the pipeline network of 110kV substations in old urban areas, the method of this invention makes adaptive adjustments to the aforementioned steps to ensure effective operation even under incomplete data conditions. The specific implementation process is as follows:

[0222] In step S1, a substitution strategy is used to obtain the necessary parameters.

[0223] First, the underground pipeline reflection profile is obtained by scanning the 0-2 meter depth of the station area using ground penetrating radar (GPR). The hyperbolic reflection characteristics of the pipeline are picked out from the profile image manually or by using an automatic recognition algorithm. The travel time of the reflected wave is recorded. Combined with the empirical value of the radar wave velocity in the soil, such as 0.1 meters per nanosecond, the burial depth of the pipeline is estimated.

[0224] Connecting the reflection points on the plan to determine the pipeline route, while using a handheld GPS device to record the planar coordinates of the manhole covers, and matching the manhole cover positions with the detected pipeline positions, thereby determining the planar coordinates and inner bottom burial depth of each discharge outlet.

[0225] Then, a pressure sensor is temporarily installed in the inspection well to continuously monitor the water level change for 24 hours and record the water level change curve over time. The lowest value of multiple low tides or stable water levels within 24 hours is taken as the approximate value of the bottom elevation of the discharge outlet. At the same time, the surface elevation at this location is combined with the surface elevation, which can be extracted from the three-dimensional model or obtained by on-site leveling measurement. The formula for calculating the bottom elevation of the discharge outlet is: the bottom elevation equals the surface elevation minus the bottom burial depth, with the unit being meters, relative to the 1985 National Elevation Datum.

[0226] For the proportion of impermeable subsurface surfaces, due to the lack of high-precision UAV imagery, Gaofen-7 satellite imagery was used instead, with a resolution of 0.8 meters and including four bands: red, green, blue, and near-infrared.

[0227] The image is segmented into image patches that match the station area, and then input into a pre-trained convolutional neural network (CNN) model for classification. This CNN model is fine-tuned based on publicly available remote sensing image classification datasets such as DeepGlobe, and outputs the surface material category for each pixel, including asphalt, concrete, lawn, and bare soil. After classification, the pixel proportion of each category is calculated to obtain the proportion of impermeable underlying surface.

[0228] Because the satellite imagery has a low resolution, the classification accuracy is about 88%. Therefore, a 5% uncertainty weight is introduced into the subsequent water balance equation. Specifically, in step S204, the baseline value φ0 of the dynamic runoff coefficient is multiplied by 1.05 to account for the underestimation of runoff that may be caused by satellite imagery classification errors.

[0229] In step S2, since there is no historical evaporation data, the initial soil water deficit depth I0 is calculated using a simplified formula: I0 = 10 + 2 × D dry ;

[0230] Among them, D dry The calculation is based on the number of consecutive rainless days in the preceding period, with an upper limit of 30 mm.

[0231] To address the issue of unknown siltation levels in the pipes and channels, a siltation coefficient β is introduced, with an initial value of 0.7, representing that 70% of the theoretical drainage capacity can be achieved, and the net drainage volume Q. net The calculation formula is adjusted as follows:

[0232] ;

[0233] Where Q(t) i The value represents the theoretical drainage capacity calculated in step S205 using the orifice outflow formula, before deducting seepage losses, and is expressed in cubic meters per hour. rate This refers to the rate of seepage loss in the pipeline network, expressed in cubic meters per meter per hour; L pipe The total length of the drainage pipes and channels is expressed in meters.

[0234] The β value will be manually corrected by the maintenance personnel in subsequent step S5 based on the actual dredging records.

[0235] In step S3, the 3D model is modeled using oblique photogrammetry as a single source, without relying on BIM data. The foundation elevation of key power equipment is manually measured by a laser rangefinder, and the measurement data is uploaded to the system via a mobile app.

[0236] To improve the accuracy of surface runoff simulation, the influence of the Manning roughness coefficient *n* on flow resistance is added to the inter-grid runoff distribution in step S304. After calculating the directional slope, when it is necessary to distribute water volume weighted by multi-directional hydraulic conductivity, the Manning coefficient is used as a supplementary factor for hydraulic conductivity, and the weighting calculation formula is adjusted as follows:

[0237] ;

[0238] Among them, D dir For directional slope, K dir Let n be the soil hydraulic conductivity of the neighborhood grid. dir is the Manning roughness coefficient of the neighborhood raster.

[0239] The Manning coefficient values ​​for different underlying surface types are as follows: n is 0.25 for lawn areas, n is 0.015 for concrete areas, and empirical values ​​for other areas can be found in relevant hydrological manuals such as the textbook "Hydraulics".

[0240] In step S4, a new "sandbag" option is added to the emergency resource database. The physical parameters of the sandbags are preset as follows: a water-blocking height of 0.3 meters and a length of 1 meter per bag. When the total rated flow of the existing drainage equipment is insufficient to meet the calculated total drainage flow Q... required The system automatically calculates the required number of sandbags. The calculation steps are as follows:

[0241] First, determine the number of sandbag stacking layers L based on the predicted water depth h. h is obtained from the water depth at the location of the protective equipment in the water level field in step S307, and the unit is meters. The number of stacking layers L is equal to h divided by 0.3 and then rounded up, i.e., L=ceil(h / 0.3).

[0242] Next, determine the length D of the area to be protected. D is taken as the perimeter of the equipment foundation or set according to the actual risk area, for example, the perimeter of the area extending 1 meter outward from the equipment foundation, in meters. The number of sandbags required for each layer, N, is also determined. per_layer It equals D divided by 1 and rounded up, which is N. per_layer =ceil(D / 1).

[0243] The final total number of sandbags required is equal to L multiplied by N. per_layer .

[0244] The optimization of the sandbag placement location adds the constraint of "shortest transport path", based on the two-dimensional grid map of the station area, which is consistent with the 0.5-meter grid in step S302, with each grid marked as feasible or infeasible.

[0245] Infeasible grids include grids within the prohibited work area polygon and pre-marked narrow passages, such as those less than 1.5 meters wide. Starting from the grid where the sandbag storage point is located and ending at the grid where the deployment point is located, the shortest feasible path is searched using the A* algorithm. The heuristic function is Euclidean distance. All infeasible grids are avoided during the search process. The output path is a recommended transport route for maintenance personnel to refer to.

[0246] In step S5, a new "Siltation Level" slider is added to the parameter correction panel. The slider value ranges from 0 to 1 for the β value, with a step size of 0.01. Maintenance personnel can adjust the slider based on on-site dredging records or experience. After adjustment, the system saves the current β value and synchronously updates the β value used in step S205. However, it should be noted that the seepage loss rate (seepage_rate) is unrelated to the siltation level, and therefore is not automatically updated here. Immediately after the slider is adjusted, a recalculation of the entire station's water volume is triggered, and steps S2 to S4 are re-executed to assess the impact of the corrected drainage capacity on the evolution of water accumulation.

[0247] After the adjustment is completed, the system automatically generates a pipeline health assessment report. The report includes the current β value and its corresponding siltation level rating. For example, β greater than 0.8 indicates slight siltation, 0.5 to 0.8 indicates moderate siltation, and less than 0.5 indicates severe siltation. It also displays the historical trend of seepage loss rate and provides pipeline maintenance or renovation recommendations based on this data, such as dredging or pipe section replacement. The report can be viewed through a 3D visualization interface for reference in subsequent operation and maintenance decisions.

[0248] Through the aforementioned targeted adjustments, the method of this invention can still complete core functions such as multi-dimensional factor library construction, dynamic water balance simulation, three-dimensional water accumulation mapping, emergency resource matching, and online model correction in the context of old substations with missing pipeline data and limited data accuracy, providing a flexible and effective solution for substation flood monitoring.

[0249] Example 3

[0250] For coastal substations with high groundwater levels, the method of this invention adds relevant parameters and correction mechanisms for the impact of groundwater to the conventional implementation method, in order to adapt to the special constraints on the water accumulation process when the groundwater level is high. The specific implementation process is as follows:

[0251] In step S, in addition to conventional topographic, pipeline, and environmental parameters, groundwater level monitoring data is collected in the station area. Groundwater observation wells are installed within the station area, with submersible water level gauges installed inside. The water level gauges have an accuracy of ±1 cm, and the sampling frequency is once per hour to monitor the groundwater level elevation H in real time. groundwater The unit is meters, relative to the 1985 National Elevation Datum. Groundwater level data is uploaded to the system database every hour via a wireless transmission module, stored in the groundwater parameter table, and spatially correlated with substation zones. Groundwater back pressure parameters are also added for subsequent water balance calculations.

[0252] In step S2, the surface infiltration loss I(t) is... i The calculation of ) adds a back pressure correction term, when the groundwater level elevation H groundwaterWhen the water depth exceeds the initial soil deficit depth I0, the infiltration capacity will be weakened by groundwater backwater. The corrected formula for calculating infiltration loss is as follows:

[0253] ;

[0254] Where I0 is the initial soil deficit water depth calculated in step S203, in millimeters, and H groundwater The current groundwater level elevation is in meters. Multiply by 1000 to convert meters to millimeters to ensure unit consistency.

[0255] The above formula ensures that when H groundwater When ×1000 is greater than or equal to I0, the infiltration loss is zero, meaning the soil is completely saturated and can no longer infiltrate; when H groundwater When ×1000 is less than I0, the infiltration loss is equal to the difference between the two, which represents the amount of water that the soil can still absorb after deducting the backwater effect. This effectively prevents negative infiltration and makes the water balance equation more consistent with the actual hydrological process in areas with high groundwater levels.

[0256] In step S3, the groundwater level contour layers are integrated into the 3D model. The contour data are generated by the kriging interpolation method using the water level gauge measurements from each observation well. A spherical variogram model is used for interpolation, with a nugget constant of 0.001 m², a sill value of 0.01 m², and a range set to 50 m based on the well spacing, forming a continuous groundwater level surface H covering the entire station area. gw (x,y), in meters, is a surface superimposed on a ground elevation model to determine whether water accumulation is constrained by groundwater.

[0257] In determining the boundary of water accumulation, the algorithm incorporates Darcy's law module to calculate the lateral seepage exchange between the accumulated water and groundwater. Specifically, when the water surface elevation H of the water accumulation grid... water (x,y,t) is greater than the groundwater level H. gw When (x, y), water will seep downwards through soil pores, and the amount of seepage is calculated using Darcy's law:

[0258] ;

[0259] Among them, Q seep K represents the lateral seepage flow rate within time step Δt, in cubic meters. sat The saturated hydraulic conductivity of the soil is obtained by converting the soil hydraulic conductivity K. K is in millimeters per hour; to convert it to meters per second, multiply by a factor of 0.001 / 3600, i.e., K. sat =K×0.001 / 3600, in meters per second; i is the hydraulic gradient, which can be approximated by dividing the head difference between the water accumulation grid and the adjacent unsaturated area by the seepage path length (H water -H gw) / d, where d is the grid size (0.5 meters); A interface The contact area between the water-accumulating grid and the unsaturated area is approximately equal to the grid side length of 0.5 meters multiplied by the water depth d of that grid. i d i =V i / S grid V i S represents the water volume accumulated in the grid. grid The grid area is 0.25 square meters.

[0260] Calculated Q seep The water volume is deducted from the water accumulation grid, and the corresponding water enters the groundwater system and is reflected in the groundwater level model, so that the water accumulation evolution results are more consistent with the actual overflow characteristics of areas with high groundwater levels.

[0261] In step S4, the Emergency Response Resource Library adds an anti-backflow gate option. The gate is usually installed at the outlet of the drainage pipe or at the point of connection with the external water body to prevent backflow of high tide or flood.

[0262] Real-time monitoring of the effective drainage pressure difference ∆H=H at the current time step water (t i )-H invert When ΔH≤0, it means that the water level at the discharge outlet is lower than or equal to the inner bottom elevation, which may cause backflow. The location of the gate that needs to be closed will be automatically highlighted in the three-dimensional interface, and the closing operation will be recommended.

[0263] Simultaneously, based on the gate size, operation method, and distance from the current personnel location, the required manpower and time are calculated. Manpower estimation is based on gate type: for manual gates, two people are required to operate each gate; for electric gates, one person is needed to start them. Time estimation is based on the shortest feasible path length between the gate location and the maintenance personnel's current location, divided by the average travel speed, which is taken as 1.5 meters per second. If multiple people are required to work together, the maximum time for each path is used.

[0264] The above calculation results are pushed to the operation and maintenance personnel's terminal via scheduling instructions.

[0265] In step S5, a groundwater level correction slider is added to the parameter correction panel. The slider is used to adjust the overall offset of the current global groundwater level elevation. The value range is -0.5 meters to +0.5 meters, the step size is 0.01 meters, and the initial value is 0.

[0266] Maintenance personnel can adjust the slider based on on-site observations or experience. After adjustment, the system will immediately update the current groundwater level surface H. gw The offset is added to the entire (x,y) system, and the water volume of the entire station is recalculated. Steps S2 to S4 are then re-executed to assess the impact of groundwater level changes on water accumulation evolution and emergency response plans.

[0267] In addition, the system establishes an API interface with the local water bureau's monitoring platform, automatically downloading the latest groundwater monitoring data for the substation area every hour, including groundwater level and trends. The downloaded data is then fused with the data from the system's observation wells, and a weighted average method is used to generate a more reliable real-time groundwater level surface. The fusion weights are dynamically set based on the confidence level of the data source. For example, if the confidence level of the water bureau's data is 0.8 and the confidence level of the station's observation data is 0.9, the fused groundwater level value is calculated as (0.8 × water bureau data + 0.9 × station data) divided by (0.8 + 0.9).

[0268] The merged groundwater level data is automatically updated to the 3D model, realizing the fusion and correction of multi-source data.

[0269] With the addition of the aforementioned targeted parameters and mechanisms, the method of this invention can accurately characterize the impact of groundwater on infiltration and waterlogging in the scenario of coastal substations with high groundwater levels, provide emergency suggestions for preventing backflow, and achieve linkage correction with external water data, significantly improving the adaptability and reliability of flood monitoring.

[0270] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments; the embodiments and descriptions in the specification are merely preferred embodiments of the present invention. Various changes and modifications can be made to the present invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent monitoring of flooding in substations, characterized in that, Includes the following steps: Construct a multi-dimensional influencing factor library, which includes geomorphic parameters, pipeline physical parameters, and environmental water consumption parameters for each functional zone of the substation. Among them, the pipeline physical parameters include at least the inner bottom elevation of each discharge outlet and the seepage loss rate of the pipeline. Acquire rainfall forecast data and combine it with a multi-dimensional influencing factor library to perform dynamic water balance simulation. This simulation process includes: calculating the initial soil deficit depth based on the number of consecutive rainless days in the previous period. The initial soil deficit depth is determined based on the soil saturation water content and the historical evaporation corresponding to the number of consecutive rainless days in the previous period. The runoff coefficient at the beginning of rainfall is dynamically corrected based on the initial soil deficit depth. The dynamic correction includes: when the cumulative rainfall is less than the initial soil deficit depth, increasing the runoff coefficient through a preset power function, without exceeding a preset upper limit; at the same time, calculating the net drainage volume based on the effective pressure difference between the current water level and the bottom elevation of the discharge outlet, combined with the seepage loss rate of the pipeline. Based on the net increase in water volume obtained from dynamic water balance simulation, combined with the high-precision three-dimensional digital model of the substation, three-dimensional water accumulation mapping is performed, and a water level height field that evolves over time is generated according to the net increase in water volume. By comparing the water level field with the foundation elevation of key power equipment, equipment at risk of flooding is identified. Based on the current net increase in water volume, the total emergency drainage demand is calculated. A combination of drainage equipment that meets this total demand is matched from the emergency resource database. Simultaneously, using the centroid of water accumulation as the target point and combining it with the preset prohibited operation area, the deployment coordinates of each drainage device are optimized. The centroid of water accumulation is obtained by calculating the weighted average coordinates of the water accumulation volume of each grid within the water accumulation area. Optimizing the deployment coordinates of each drainage device includes: using the centroid of water accumulation as the target, calculating the Euclidean distance between each feasible deployment point in the station area (excluding the preset prohibited operation area) and the target, and selecting the point with the smallest distance as the deployment coordinates of the device. A dispatch instruction containing the deployment coordinates and the estimated arrival time is generated.

2. The intelligent flood monitoring method for substations according to claim 1, characterized in that, The construction steps of the multidimensional impact factor library include: The total area of ​​the station area was extracted by the geographic information system, the local slope ratio was obtained by gradient calculation through the digital elevation model, the proportion of impermeable underlying surface was obtained by fusing lidar point cloud and UAV oblique photography images and using support vector machine classification, and the soil hydraulic conductivity was obtained by field measurement or database call and the confidence level was marked. The inner bottom elevation of each discharge outlet is extracted by total station measurement or building information model. The maximum depth of the pipeline is determined by three-dimensional pipeline model combined with on-site verification. The seepage loss rate of the pipeline is determined by referring to the specifications or water injection test based on the pipe material type and service life. Vegetation cover type was retrieved by inverting multispectral images from drones, and the average daily water consumption per unit area was calculated by combining the reference crop evapotranspiration and crop coefficient provided by the weather station.

3. The intelligent flood monitoring method for substations according to claim 1, characterized in that, In the dynamic water balance simulation, the steps for dynamically correcting the runoff coefficient include: The runoff coefficient is set to a baseline value equal to the proportion of the impermeable underlying surface. Determine whether the cumulative rainfall is less than the initial soil water deficit depth. If so, increase the runoff coefficient according to an increasing function, provided it does not exceed a preset upper limit. The increasing function is a power function, and its form is: If not, the runoff coefficient is restored to the baseline value; The initial water deficit depth of the soil is calculated based on the soil saturation water content and the historical evaporation corresponding to the number of consecutive rainless days in the previous period.

4. The intelligent flood monitoring method for substations according to claim 1, characterized in that, The calculation steps for the net drainage volume include: Traverse all discharge outlets within the station area, select the discharge outlet with the lowest inner bottom elevation that is lower than the current water level as the effective drainage outlet, and calculate the difference between its inner bottom elevation and the current water level as the effective pressure difference; When the effective pressure difference is greater than zero, the theoretical drainage capacity is calculated based on the orifice outflow principle. Then, the product of the seepage loss rate of the pipeline and the total length of the drainage pipeline is deducted to obtain the net drainage volume.

5. The intelligent flood monitoring method for substations according to claim 1, characterized in that, The steps of the three-dimensional water accumulation mapping include: The surface of the station area is divided into regular grids, and each grid records the ground elevation, slope and soil hydraulic conductivity; The net increase in water volume at each time step is evenly distributed to each grid to obtain the initial water accumulation volume; The D8 flow model is used to distribute water flow based on the ground slope and soil hydraulic conductivity between each grid cell. The process is iterated until a steady state is reached, and the final water accumulation volume of each grid cell is obtained. Based on the water volume of each grid, the centroid coordinates of the water accumulation area are calculated, and the water spread boundary is determined by the flood filling algorithm, ultimately generating the water level height field.

6. The intelligent flood monitoring method for substations according to claim 1, characterized in that, The steps for matching and optimizing the layout coordinates of the drainage equipment combination include: Available equipment is selected from the emergency resource pool and sorted by rated flow rate. A greedy algorithm is then used to select the minimum equipment combination that meets the required flow rate. The station area ground is discretized into a grid. Using the coordinates of the water accumulation centroid as the target, the Euclidean distance from all feasible grid points to the centroid is calculated. After excluding grid points that fall into the prohibited operation area, the grid point with the smallest distance is selected as the layout coordinate. The straight-line distance is calculated based on the coordinates of the equipment storage point and the deployment coordinates. The estimated arrival time is estimated by combining the average travel speed. At the same time, a path planning algorithm is used to generate a recommended travel path from the storage point to the deployment point.

7. The intelligent flood monitoring method for substations according to claim 1, characterized in that, It also includes steps for online model calibration and evolution: Real-time access to ultrasonic water level sensor data deployed in the station area; calculation of root mean square error between simulated water depth and measured water depth. When the error exceeds the preset threshold, a manual interactive parameter correction slider is provided in the 3D visualization interface to adjust the local soil hydraulic conductivity or the global pipe and canal seepage loss rate. Upon receiving the confirmation instruction, the current simulation is paused, the corrected parameters are reloaded, and the recalculation of the station's water accumulation evolution is initiated from the moment the rainfall begins, updating the water level field and emergency dispatch plan.

8. The intelligent flood monitoring method for substations according to claim 7, characterized in that, The recalculation process is accelerated using parallel processing with a graphics processor: The station area grid is divided into multiple sub-regions, each of which is assigned to an independent GPU thread block. The sub-region data is cached using shared memory for parallel computation, and the water accumulation evolution prediction for the next 2 hours is completed within 30 seconds.

9. The intelligent flood monitoring method for substations according to claim 1, characterized in that, For different specific scenarios, at least one of the following adaptive adjustments is also included: For old substations with missing pipeline data, ground-penetrating radar is used to detect the direction and depth of the pipelines, and pressure sensors are used to monitor and estimate the bottom elevation of the discharge outlet. At the same time, a siltation coefficient is introduced to correct the net drainage volume. For coastal substations with high groundwater levels, groundwater level monitoring data is collected, groundwater backwater correction is introduced into the calculation of surface infiltration loss, and lateral seepage calculation based on Darcy's law is added in the process of determining the water spread boundary. For scenarios with insufficient emergency resources, sandbags are added as emergency resources. The number of sandbags is automatically calculated based on the predicted water depth and the length of the area to be protected, and the shortest transportation route is planned.

10. A substation flood intelligent monitoring system, characterized in that, include: The multidimensional influencing factor library construction module is used to acquire and store the geomorphological parameters, pipeline physical parameters and environmental water consumption parameters of each functional zone of the substation. The pipeline physical parameters include at least the inner bottom elevation of each discharge outlet and the pipeline seepage loss rate. The dynamic water balance extrapolation module is used to perform time-discrete water conservation calculations based on rainfall forecast data and the multi-dimensional influencing factor library. This module includes: a soil initial deficit water depth calculation unit, used to calculate the soil initial deficit water depth based on the number of consecutive rainless days in the previous period. The soil initial deficit water depth is determined based on the soil saturated water content and the historical evaporation corresponding to the number of consecutive rainless days in the previous period. Based on the soil initial deficit water depth, the runoff coefficient at the beginning of rainfall is dynamically corrected. The dynamic correction includes: when the cumulative rainfall is less than the soil initial deficit water depth, the runoff coefficient is increased by a preset power function, which does not exceed a preset upper limit; at the same time, the net drainage volume is calculated based on the effective pressure difference between the current water level and the bottom elevation of the discharge outlet, combined with the seepage loss rate of the pipeline. The three-dimensional water accumulation mapping module is used to generate a water level height field that evolves over time based on the net increase in water volume obtained through deduction and combined with the high-precision three-dimensional digital model of the substation. The emergency resource scheduling module is used to determine the risk of equipment flooding based on the water level field and the foundation elevation of key power equipment, calculate the total emergency drainage demand based on the net increase in water volume, match drainage equipment combinations that meet the demand from the emergency resource database, and optimize equipment deployment coordinates by using the water accumulation centroid as the target point and combining it with the prohibited operation area. The water accumulation centroid is obtained by calculating the weighted average coordinates of the water accumulation volume of each grid in the water accumulation area. Optimizing the deployment coordinates of each drainage device includes: using the water accumulation centroid as the target, calculating the Euclidean distance between each feasible deployment point in the station area (excluding the preset prohibited operation area) and the target, selecting the point with the smallest distance as the deployment coordinates of the device, and generating a scheduling instruction that includes the deployment coordinates and the estimated arrival time. The online model calibration module is used to access sensor measured data in real time, receive parameter correction instructions through the human-computer interaction interface, and trigger the recalculation of the water accumulation evolution of the entire station to update the water level field and scheduling scheme.

Images