Flood prevention unmanned aerial vehicle scheduling method based on digital elevation and meteorological model

Abstract

The present application relates to the technical field of unmanned aerial vehicle cluster scheduling and flood control emergency response, and particularly relates to a flood control unmanned aerial vehicle scheduling method based on digital elevation and meteorological model, which has the technical scheme as follows: DEM data and HRES data of a target river basin are acquired; a flood evolution hydrodynamic model is constructed based on a two-dimensional shallow water wave equation to predict the time-space evolution path of a submerged range, water depth and flow velocity; high-risk dynamic nodes are extracted accordingly, and the best hovering observation viewpoint meeting the requirements of no terrain obstruction and local wind speed safety is calculated in combination with DEM and HRES wind field; a multi-objective optimization cost function including observation benefits, flight energy consumption and meteorological risks is constructed, and an improved particle swarm or ant colony algorithm is used to solve an optimal scheduling scheme to dynamically allocate a patrol route and a hovering observation point for an unmanned aerial vehicle cluster; the present application realizes predictive dynamic scheduling driven by flood evolution, reduces the flight energy consumption of the cluster, and improves the efficiency of flood control emergency response.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) swarm scheduling and flood control emergency response technology, and in particular to a flood control UAV scheduling method based on digital elevation and meteorological models. Background Technology

[0002] Currently, the dispatching of drones for flood control emergency response mainly relies on static key monitoring lists or temporary manual assignments, lacking quantitative prediction of the dynamic evolution of floods. Regarding flight path planning, existing methods typically only consider the shortest path in a two-dimensional plane, failing to deeply integrate high-resolution meteorological forecast data (such as wind fields and rainfall intensity) with three-dimensional terrain data (DEM) to assess flight energy consumption and safety. This makes drones highly susceptible to risks such as terrain signal obstruction, crashes due to strong winds, or being unable to return due to battery depletion caused by headwinds when performing flood control missions in complex mountainous areas or under severe weather conditions.

[0003] A search revealed several related technologies disclosed in existing patent documents. For example, patent document CN108981714B discloses a method for UAV path planning and feasibility analysis. This method generates initial flyable paths by constructing Thiessen polygons and uses an ant colony algorithm to search for the shortest feasible path, effectively avoiding areas threatened by static obstacles. However, this method's path planning is based solely on obstacle avoidance in a two-dimensional plane. It fails to couple high-resolution weather forecasts (such as HRES wind fields and rainfall intensity) with three-dimensional terrain DEMs to calculate flight energy consumption costs, nor does it introduce flood hydrodynamic models to drive the generation of dynamic target points. Therefore, it cannot achieve predictive dynamic route optimization for flood control and emergency response tasks. Another example is patent document CN114330825B, which discloses an integrated method for flood forecasting and scheduling in large river basins. Through node-based generalization and categorized sequential calculation, it achieves joint calculation of flood forecasting within the basin with the scheduling of ground infrastructure such as reservoirs and river cross-sections. However, this method only applies to ground-based water conservancy facilities and does not integrate aerial patrol resources of UAV swarms. It cannot directly drive UAV swarms to complete accurate waypoint planning and task allocation by using high-risk dynamic nodes predicted by hydrodynamic models.

[0004] In summary, existing technologies lack a comprehensive algorithmic framework that deeply couples hydrodynamic models, high-resolution meteorological models, and UAV operations research scheduling. Specifically, existing scheduling systems cannot automatically extract high-risk monitoring nodes based on dynamic predictions of flood evolution, nor can they simultaneously consider terrain occlusion and meteorological wind field constraints in three-dimensional space to optimize the hovering observation viewpoint of UAVs. Furthermore, they lack a multi-objective cost function that integrates flood risk weights, meteorological energy consumption models, and spatial distance to achieve dynamic redistribution of UAV swarm tasks. Therefore, there is an urgent need to propose a UAV scheduling method capable of achieving "predictive and precise deployment" of flood control emergency resources.

[0005] In view of this, we propose a flood control drone scheduling method based on digital elevation and meteorological models to solve the existing problems. Summary of the Invention

[0006] The purpose of this invention is to provide a flood control drone scheduling method based on digital elevation and meteorological models to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a flood control drone scheduling method based on digital elevation and meteorological models, comprising the following steps:

[0008] S1: Acquire high-precision digital elevation model (DEM) data and high-resolution weather forecast (HRES) data for the target watershed. The HRES data includes the spatiotemporal distribution of rainfall and three-dimensional wind field data.

[0009] S2: Based on rainfall data from DEM and HRES data, a two-dimensional shallow water wave equation is used to construct a flood evolution hydrodynamic model to predict the spatiotemporal evolution path of the flood in the watershed in the future period. The spatiotemporal evolution path includes the inundation range, water depth and flow velocity.

[0010] S3: Based on the spatiotemporal evolution path of the flood predicted in step S2, extract one or more high-risk dynamic nodes, and for each high-risk dynamic node, calculate the optimal hovering observation point that satisfies the conditions of no terrain obstruction and local wind speed below the wind resistance limit of the UAV, in combination with DEM data.

[0011] S4: Construct a multi-objective optimization cost function, which includes an observation benefit term, a flight energy consumption term affected by HRES 3D wind field data, and a meteorological risk penalty term;

[0012] S5: Utilize optimization algorithms to solve the multi-objective optimization cost function, dynamically allocate patrol routes and hovering observation points for the UAV swarm, and generate an optimal scheduling scheme that includes the three-dimensional flight paths and task points of each UAV.

[0013] Furthermore, the optimization algorithm in step S5 is an improved particle swarm optimization algorithm or an improved ant colony optimization algorithm.

[0014] Furthermore, in step S2, the finite volume method is used to spatially discretize and solve the two-dimensional shallow water wave equation, and a graphics processing unit (GPU) is used to accelerate the calculation.

[0015] Furthermore, step S1 also includes a preprocessing step for the HRES data: using cubic spline interpolation to match the grid resolution of the HRES data to the grid resolution of the DEM data, and using Gaussian filtering to smooth the interpolation results for steep slope areas in mountainous regions.

[0016] Furthermore, the specific method for calculating the optimal hovering observation viewpoint in step S3 is as follows: under the conditions that the line of sight from the candidate hovering point to the high-risk dynamic node is unobstructed from the DEM data and the local wind speed at the hovering point is less than the wind resistance limit of the UAV, the spatial coordinate with the minimum local wind speed is selected as the optimal hovering observation viewpoint.

[0017] Furthermore, the multi-objective optimization cost function J in step S4 is expressed as: ,in, Let i be the risk weight of the i-th high-risk dynamic node. The flight energy consumption of the j-th UAV affected by the wind field. Let α, β, and γ be the weather risk penalty term encountered by the j-th drone, and α, β, and γ be the weighting coefficients.

[0018] Furthermore, the specific method for extracting high-risk dynamic nodes in step S3 is as follows: calculate the water depth change rate dh / dt of each grid cell in the flood evolution hydrodynamic model, and extract the grid cells whose water depth change rate exceeds the preset threshold as high-risk dynamic nodes.

[0019] Furthermore, it also includes a real-time replanning step: when the rate of water level change in any grid cell exceeds the preset replanning threshold, only the drone task allocation scheme within the affected radius is locally adjusted.

[0020] Furthermore, the local wind speed at the optimal hovering observation point in step S3 is obtained by combining HRES three-dimensional wind field data with DEM topographic gradient for micro-topographic wind field correction.

[0021] Furthermore, when using the improved ant colony algorithm, the pheromone evaporation coefficient ρ is dynamically adjusted according to the real-time meteorological risk level: when the regional rainfall intensity is greater than 50 mm / h, the pheromone evaporation coefficient ρ is increased to 0.25.

[0022] Compared with the prior art, the beneficial effects of the present invention are:

[0023] 1. This invention breaks through the limitations of traditional flood control drone scheduling, which relies on static lists or manual assignments. For the first time, it deeply couples a flood evolution hydrodynamic model based on two-dimensional shallow water wave equations with drone swarm scheduling. By inputting HRES rainfall data as the source term into the shallow water wave equations and combining it with DEM topographic gradients, it can quantitatively predict the three-dimensional spatiotemporal evolution path of flood inundation range, water depth, and flow velocity in future periods, and automatically extract high-risk dynamic nodes where the water depth change rate exceeds a threshold. This "predictive scheduling" mechanism shifts drone task allocation from passive response to proactive prediction, avoiding ineffective patrols in areas where floods have not yet arrived or have receded. In practical applications in the Yellow River Basin, the flood inundation range prediction error was reduced from ±25% using traditional methods to ±12%, and the system successfully controlled the average response time of 37 drones to within 8 seconds. In an application on a tributary in the middle reaches of the Yangtze River, the system provided an early warning of the risk of dike breach at a critical section 47 minutes in advance, buying valuable time for emergency decision-making. It achieves predictive dynamic scheduling driven by flood evolution, significantly improving the efficiency of flood control emergency response.

[0024] 2. This invention addresses the challenges of strong winds and complex mountainous terrain often encountered at flood control sites by proposing a method for calculating the optimal hovering observation viewpoint that combines terrain obstruction analysis with micro-topographic wind field correction. Ray tracing is used to determine whether DEM terrain obstruction exists between candidate hovering points and high-risk target nodes. Simultaneously, HRES 3D wind field data and the DEM terrain gradient function are used to calculate local wind speeds. The point with the minimum local wind speed that satisfies the constraints of no obstruction and safe distance is selected as the optimal hovering coordinate. This scheme effectively avoids strong wind zones caused by the "tunneling effect" in valleys and terrain signal obstruction, enabling UAVs to hover stably in safe areas such as leeward slopes and acquire clear images of flood evolution. Experimental data shows that the UAV mission completion rate has increased from 68% in existing technologies to 93%, and the rate of crashes and loss of contact due to strong winds or obstruction has significantly decreased; ensuring UAV flight safety and data acquisition quality under severe weather and complex terrain conditions.

[0025] 3. This invention constructs a multi-objective optimization cost function, in which the flight energy consumption term explicitly incorporates the nonlinear influence of the HRES wind field on the UAV's propulsion power. Optimized flight path planning guides the UAV to fly with the wind as much as possible and avoid strong wind shear areas, thereby significantly reducing the additional energy consumption caused by headwind flight. Simultaneously, an improved particle swarm optimization or ant colony optimization algorithm is used to solve for the optimal task allocation and flight path matrix, ensuring that the swarm minimizes overall energy consumption while covering all high-risk nodes. Experimental results show that path planning energy consumption is reduced by 28%, and the single-inspection endurance time is extended by 40 minutes, effectively alleviating the pain point of insufficient UAV endurance in flood control emergencies; reducing swarm flight energy consumption and extending single-inspection endurance time.

[0026] 4. This invention also includes a real-time replanning mechanism: when the water level change rate |dh / dt| of any grid cell is detected... grid When the speed exceeds 0.15 m / h, the system automatically initiates local replanning, dynamically adjusting only the drone missions within a 2km radius of the affected area to avoid the computational overload of global recalculation. This mechanism enables the algorithm to maintain its global optimization capabilities while rapidly responding to emergencies such as sudden water level surges and dam breaches. In a cluster experiment involving 37 drones, the average response time remained below 8 seconds, demonstrating excellent real-time performance and scalability. It achieved real-time replanning, balancing global optimization with timely local responses.

[0027] 5. This invention organically combines multi-source data fusion, finite volume method for solving shallow water wave equations, and multi-objective optimization scheduling to form a complete closed-loop algorithm framework encompassing data input, flood evolution prediction, high-risk node extraction, and UAV task allocation and flight path planning. This invention is the first to deeply integrate hydrodynamic models, high-resolution meteorological models, and UAV swarm operations research scheduling, achieving "precise deployment" and "dynamic optimization" of flood control resources. It has made substantial progress in prediction accuracy, flight safety, energy consumption optimization, and real-time response, comprehensively improving the intelligence level of basin flood control emergency response. Attached Figure Description

[0028] Figure 1 This is a one-way flowchart of a flood control drone scheduling method based on digital elevation and meteorological models according to the present invention.

[0029] Figure 2 This is a core method framework diagram of a flood control drone scheduling method based on digital elevation and meteorological models according to the present invention;

[0030] Figure 3 This is a flood inundation range prediction map for a flood control drone scheduling method based on digital elevation and meteorological models according to the present invention. Detailed Implementation

[0031] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0032] Example 1

[0033] A flood control drone scheduling method based on digital elevation and meteorological models includes the following steps:

[0034] S1: Acquire high-precision digital elevation model (DEM) data and high-resolution weather forecast (HRES) data for the target watershed. The HRES data includes the spatiotemporal distribution of rainfall and three-dimensional wind field data.

[0035] S2: Based on rainfall data from DEM and HRES data, a two-dimensional shallow water wave equation is used to construct a flood evolution hydrodynamic model to predict the spatiotemporal evolution path of the flood in the watershed in the future period. The spatiotemporal evolution path includes the inundation range, water depth and flow velocity.

[0036] S3: Based on the spatiotemporal evolution path of the flood predicted in step S2, extract one or more high-risk dynamic nodes, and for each high-risk dynamic node, calculate the optimal hovering observation point that satisfies the conditions of no terrain obstruction and local wind speed below the wind resistance limit of the UAV, in combination with DEM data.

[0037] S4: Construct a multi-objective optimization cost function, which includes an observation benefit term, a flight energy consumption term affected by HRES 3D wind field data, and a meteorological risk penalty term;

[0038] S5: Utilize optimization algorithms to solve the multi-objective optimization cost function, dynamically allocate patrol routes and hovering observation points for the UAV swarm, and generate an optimal scheduling scheme that includes the three-dimensional flight paths and task points of each UAV.

[0039] Step S1 also includes a preprocessing step for the HRES data: using cubic spline interpolation to match the grid resolution of the HRES data to the grid resolution of the DEM data, and smoothing the interpolation results for steep slope areas in mountainous regions using Gaussian filtering. Step S2 uses the finite volume method to spatially discretize and solve the two-dimensional shallow water wave equations, and utilizes a graphics processing unit (GPU) for accelerated computation. The optimization algorithm in step S5 is either an improved particle swarm optimization algorithm or an improved ant colony optimization algorithm.

[0040] The HRES meteorological model, also known as high-resolution weather forecasting, is a high-resolution deterministic weather forecasting model that provides detailed predictions of meteorological elements such as precipitation and wind fields. Flood evolution models, also known as inundation analysis models or hydrodynamic models, are mathematical models that simulate the flow, diffusion, and inundation processes of floods in complex terrain (DEM) based on fluid dynamics equations (such as the two-dimensional shallow water wave equation). HRES, also known as the HRES system, is a high-resolution numerical model of the European Centre for Medium-Range Weather Forecasts. A DEM is a digital elevation model that describes a three-dimensional topographic dataset of a geographic region.

[0041] Figure 1 It demonstrates the entire process from data input, flood evolution prediction, high-risk node extraction to drone swarm scheduling. Figure 2 The system architecture, consisting of a data layer, an algorithm layer, and an execution layer, is demonstrated. The data layer includes DEM terrain data, HRES weather forecasts, and UAV swarm status. The algorithm layer includes flood evolution models (shallow water wave equations), high-risk node extraction, and multi-objective optimization scheduling (PSO / ant colony). The execution layer includes a task allocation matrix and a 3D patrol route.

[0042] By deeply coupling flood evolution prediction with UAV scheduling, a leap from "blind inspection" to "predictive and precise monitoring" has been achieved. Through implementation in a pilot section of the Yellow River Basin, the prediction error of flood inundation area has been reduced from ±25% using traditional methods to ±12%. Optimal hovering observation points are calculated by combining DEM and HRES wind field data, ensuring flight safety and image acquisition quality for UAVs under severe weather conditions. The system successfully increased the UAV mission completion rate from 68% to 93%. The multi-objective optimization scheduling algorithm maximizes the overall observation benefits of the cluster, effectively improving the efficiency of flood control emergency response in the basin. Path planning energy consumption is reduced by 28%, and the single inspection endurance time is extended by 40 minutes. In one experiment, the system successfully controlled the average response time of 37 UAVs to within 8 seconds. Figure 3 It shows the three-dimensional inundation range of the flood evolution prediction.

[0043] The working principle of the flood control drone dispatching method based on digital elevation and meteorological model according to Embodiment 1 is as follows:

[0044] For the flood evolution and high-risk node extraction algorithm based on the shallow water wave equation, HRES rainfall forecast is used as the source term input to the two-dimensional shallow water wave equation. Combined with the DEM topographic gradient, the changes in water depth and flow velocity over time at each grid point are calculated. Grids with a water depth change rate (dh / dt) exceeding a threshold are extracted as high-risk dynamic nodes. The finite volume method is used to solve the shallow water wave equation, and GPU acceleration is utilized to ensure real-time performance in emergency response. The mathematical formula is as follows: h / t + (hu) / x + (hv) / y = R(x,y,t), (hu) / t + (hu 2 + 0.5gh 2 ) / x + (huv) / y = -gh( Z / x) - τ x / ρ, where h is the water depth, u and v are the velocity components, R is the HRES rainfall source term, Z is the DEM elevation, and τ x Let g be the bottom friction force, g be the acceleration due to gravity, and ρ be the density of water. Inputs are: a high-precision DEM matrix and an HRES spatiotemporal rainfall tensor; outputs are: a spatiotemporal flood inundation map h(x,y,t) and a set of high-risk dynamic nodes P. riskThe time step Δt = 10 minutes, and the spatial resolution is consistent with that of the DEM.

[0045] For the multi-objective cluster dynamic scheduling optimization algorithm, a cost function is constructed, comprehensively considering node risk weights, UAV flight energy consumption (affected by HRES wind field), and the safety of the observation viewpoint. An improved particle swarm optimization (PSO) algorithm is used to solve for the optimal task allocation and flight path matrix. The cost function incorporates the nonlinear influence of HRES wind field on energy consumption, guiding UAVs to fly with the wind as much as possible or avoid strong wind shear areas. The mathematical formula is as follows: , ,in, Let i be the risk weight of the i-th high-risk dynamic node. The flight energy consumption of the j-th UAV affected by the wind field. Let V be the weather risk penalty term encountered by the j-th UAV, where α, β, and γ are weighting coefficients, P is the propulsion power function, and power increases when flying against the wind. drone For the speed of the drone, V windHRES For wind speed. Input the set of high-risk dynamic nodes P. risk The system uses HRES 3D wind field data and the current status of the UAV swarm (location, battery level) to output the optimal task allocation scheme for the UAV swarm and the 3D patrol route and hovering observation point of each UAV. The PSO particle count is 100 and the maximum number of iterations is 200.

[0046] To improve the ant colony algorithm, a meteorological risk weighting factor η is introduced based on the traditional ant colony algorithm. The convergence speed is improved by dynamically adjusting the pheromone evaporation coefficient ρ (0.1-0.3). A parallel computing architecture is adopted, where each UAV node independently updates its local pheromone, and global pheromone fusion is performed after each task cycle. The mathematical formula is as follows: , , where p ij Let τ be the transition probability of a drone (or ant) ​​choosing the next node j from the current node i. ij Let η be the pheromone concentration along path (i, j). ij As a heuristic factor, allowed is the set of the next nodes that can be selected starting from the current node i, i.e., the set of feasible paths, and k is the summation index. Iterating through each node k in the allowed set, τ... ij (t) represents the pheromone concentration on path (i, j) at the t-th iteration, τ ij (t+1) represents the pheromone concentration on this path at the (t+1)th iteration (i.e., the next iteration), Δτ ijThis represents the incremental pheromone addition on this path during this iteration. Input the current node i and the candidate path set, output the optimal path sequence, with pheromone importance weight α=2, heuristic factor weight β=1.5, and initial volatile coefficient ρ=0.2. In the parameter settings of the improved ant colony algorithm, the pheromone volatile coefficient ρ is adjusted according to the real-time meteorological risk level. When the regional rainfall intensity > 50 mm / h, ρ is automatically increased to 0.25 to accelerate path updates. The formula for calculating the heuristic factor η is... , where R r The regional inundation risk index is represented by γ = 0.8, and the meteorological impact coefficient is d. ij D0 is the path distance (or flight path length) from the current node i to the next node j, and D0 is the reference distance.

[0047] For the optimal hovering observation viewpoint calculation model, in three-dimensional space, combining terrain occlusion analysis and micro-topography wind field correction, the hovering coordinates of the UAV that provide both clear target visibility and maximum safety are found. Its equation is V windlocal (x,y,z) = f(V windHRES DEM gradient ), Visibility(x,y,z,Target) = Ray Cost (DEM), Optimal point =argmin(V windlocal ) st Visibility == True AND Distance < D max , where V windlocal To incorporate the local wind speed after terrain correction, V windHRES The raw wind field data obtained from the HRES meteorological model, DEM gradient Here, represents the terrain gradient of the digital elevation model, f is the correction function, Visibility is the Boolean value for view occlusion, Target is the high-risk node, and Ray... Cost (DEM) is the ray tracing function, Optimal point For the final output of the optimal hovering viewpoint coordinates, argmin(V windlocal To select, from all candidate points satisfying the constraints, the point that allows the local wind speed V to be maximized. windlocal The smallest point, with Visibility == True as one of the constraints, requires that there be no DEM terrain occlusion between the selected hovering point and the high-risk node, and Distance < D. max As a second constraint, the horizontal distance between the hovering point and the high-risk node must be less than the preset maximum observation distance D. max The constraints are: there must be no DEM terrain obstruction between the hovering point and the target node, and the local wind speed must be less than the wind resistance limit of the UAV.

[0048] For the flood evolution prediction model, based on the Saint-Venant equations and combined with the DEM elevation gradient to calculate the flow direction, spatial discretization is performed using the finite volume method. The equations are as follows: +▽·(hu) = P grid (x,y,t) - E(x,y,t)- R(x,y,t), Where h is the water depth (m), u is the velocity vector (m / s), and P grid Where E is the precipitation rate (mm / h), E is the evaporation rate (mm / h), and R is the drainage rate (m³ / h). 3 / s / km 2 ), w i P represents the bilinear interpolation weights. i The precipitation rate is set at the HRES grid. The terrain slope constraint is that areas with a slope ≥ 0.05 trigger priority warnings; the meteorological data resolution matching is HRES grid spacing ≤ 2.5 km. In HRES meteorological data preprocessing, the original grid data (spatial resolution 2.5 km) is matched to the 2 m grid of the DEM using cubic spline interpolation to ensure the spatial correspondence between terrain and precipitation; Gaussian filtering is used to eliminate data spikes in steep mountainous areas.

[0049] When the rate of water level change in a certain grid cell exceeds a threshold, i.e. Local replanning is initiated in a timely manner, adjusting only the drone missions within an affected radius of 2km to avoid a surge in global computational load. In one experiment, the system successfully kept the average response time of 37 drones within 8 seconds.

[0050] Example 2

[0051] This example demonstrates emergency monitoring of sudden torrential rain in a mountainous watershed. A sudden downpour occurred in a mountainous watershed, and the command center needed to deploy five drones to monitor villages at risk of flash floods and riverbank overflows. The system integrated hourly rainfall forecasts from a 10-meter resolution DEM (Digital Equation Model) and HRES (Hyper-Range Resonance Model). A shallow water wave equation model quickly predicted a surge in water levels in the low-lying river sections where villages A and B were located within the next three hours (identified as high-risk nodes). The system calculated the optimal viewpoint for village A, identifying strong crosswinds directly overhead. Based on the DEM, the viewpoint was adjusted to the leeward slope, halfway up the mountain (unobstructed and with low wind speed). The PSO (Positive Surface Occlusion) algorithm, based on the drones' current location and battery level, assigned the closest and most powerful drones (numbers 1 and 2) to villages A and B. The drones flew along a planned route, avoiding strong wind shear zones, to the optimal viewpoint, transmitting clear images of the flood's evolution, thus buying valuable time for evacuation.

[0052] In mountainous flood control efforts, blindly launching drones is highly susceptible to crashing due to strong winds or failing to capture crucial footage due to terrain obstruction. The core of this implementation lies in "predictive scheduling." The system does not rely on human experience but instead allows the hydrodynamic model (shallow water wave equation) to "run" ahead of the flood based on HRES rainfall forecasts, accurately pinpointing the most dangerous areas in the coming hours. When calculating hovering observation points, the system cleverly combines the HRES wind field with the DEM (Digital Equation Model), avoiding the strong wind zones of the "funnel effect" in valleys and selecting safe areas such as leeward slopes for hovering. A multi-objective cost function ensures that the entire cluster covers all high-risk nodes with minimal energy consumption and risk, achieving intelligent and precise scheduling of flood control resources. Furthermore, the initial field of the shallow water wave equation is assimilated and calibrated using real-time water level data from hydrological stations.

[0053] Example 3

[0054] This example illustrates flood control scheduling on a tributary of the middle reaches of the Yangtze River, with a drainage area of ​​875 km². 2 It includes three key monitoring sections. DEM data with a resolution of 2m and hourly updated precipitation forecasts from HRES were acquired to establish a grid system for flood evolution prediction models (spatial step size 50m). The cost function parameters were set as w1 = 0.5, w2 = 0.3, and w3 = 0.2.

[0055] First, the gridded precipitation data from HRES was transformed to the DEM coordinate system using bilinear interpolation to ensure accurate meteorological parameters for each terrain grid cell. Flood evolution calculations employed the explicit finite volume method, with a time step set to 15 minutes to guarantee numerical stability. When the predicted water level at a monitoring section was expected to rise by more than 0.3m within 2 hours, the system automatically triggered a drone swarm to dispatch 3 additional patrol missions to the area and adjusted the endurance strategies of other drones: drones >10km from the danger zone were given priority for routine inspections; those <5km were placed on standby, ready for emergency takeoff. In a practical application, the system successfully provided a 47-minute advance warning of a potential breach risk at a critical river section, and the drone swarm collected crucial image data, providing vital information for emergency decision-making. Furthermore, when meteorological data was missing, an LSTM precipitation prediction model was used to complete the data.

[0056] The above specific embodiments are merely several preferred embodiments of the present invention. Based on the technical solutions of the present invention and the relevant teachings of the above embodiments, those skilled in the art can make various alternative improvements and combinations to the above specific embodiments.

Claims

1. A method for scheduling flood control drones based on digital elevation and meteorological models, characterized in that, Includes the following steps: S1: Acquire high-precision digital elevation model (DEM) data and high-resolution weather forecast (HRES) data for the target watershed. The HRES data includes the spatiotemporal distribution of rainfall and three-dimensional wind field data. S2: Based on rainfall data from DEM and HRES data, a two-dimensional shallow water wave equation is used to construct a flood evolution hydrodynamic model to predict the spatiotemporal evolution path of the flood in the watershed in the future period. The spatiotemporal evolution path includes the inundation range, water depth and flow velocity. S3: Based on the spatiotemporal evolution path of the flood predicted in step S2, extract one or more high-risk dynamic nodes, and for each high-risk dynamic node, calculate the optimal hovering observation point that satisfies the conditions of no terrain obstruction and local wind speed below the wind resistance limit of the UAV, in combination with DEM data. S4: Construct a multi-objective optimization cost function, which includes an observation benefit term, a flight energy consumption term affected by HRES 3D wind field data, and a meteorological risk penalty term; S5: Utilize optimization algorithms to solve the multi-objective optimization cost function, dynamically allocate patrol routes and hovering observation points for the UAV swarm, and generate an optimal scheduling scheme that includes the three-dimensional flight paths and task points of each UAV.

2. The flood control drone scheduling method based on digital elevation and meteorological model according to claim 1, characterized in that: The optimization algorithm in step S5 is either an improved particle swarm optimization algorithm or an improved ant colony optimization algorithm.

3. The flood control drone scheduling method based on digital elevation and meteorological model according to claim 1, characterized in that: In step S2, the finite volume method is used to spatially discretize and solve the two-dimensional shallow water wave equation, and the graphics processing unit (GPU) is used to accelerate the calculation.

4. The flood control drone scheduling method based on digital elevation and meteorological model according to claim 1, characterized in that, Step S1 also includes a preprocessing step for the HRES data: using cubic spline interpolation to match the grid resolution of the HRES data to the grid resolution of the DEM data, and using Gaussian filtering to smooth the interpolation results for steep slope areas in mountainous regions.

5. The flood control drone scheduling method based on digital elevation and meteorological model according to claim 1, characterized in that, The specific method for calculating the optimal hovering observation viewpoint in step S3 is as follows: under the conditions that the line of sight from the candidate hovering point to the high-risk dynamic node is unobstructed from the DEM data and the local wind speed at the hovering point is less than the wind resistance limit of the UAV, the spatial coordinate with the minimum local wind speed is selected as the optimal hovering observation viewpoint.

6. The flood control drone scheduling method based on digital elevation and meteorological model according to claim 1, characterized in that: The multi-objective optimization cost function J in step S4 is expressed as: ,in, Let i be the risk weight of the i-th high-risk dynamic node. The flight energy consumption of the j-th UAV affected by the wind field. Let α, β, and γ be the weather risk penalty term encountered by the j-th drone, and α, β, and γ be the weighting coefficients.

7. The flood control drone scheduling method based on digital elevation and meteorological model according to claim 1, characterized in that, The specific method for extracting high-risk dynamic nodes in step S3 is as follows: calculate the water depth change rate dh / dt of each grid cell in the flood evolution hydrodynamic model, and extract the grid cells whose water depth change rate exceeds the preset threshold as high-risk dynamic nodes.

8. The flood control drone scheduling method based on digital elevation and meteorological model according to claim 1, characterized in that, It also includes a real-time replanning step: when the rate of water level change in any grid cell exceeds the preset replanning threshold, only the drone mission allocation scheme within the affected radius is locally adjusted.

9. The flood control drone scheduling method based on digital elevation and meteorological model according to claim 1, characterized in that: The local wind speed at the optimal hovering observation point in step S3 is obtained by combining HRES 3D wind field data with DEM topographic gradient for micro-topographic wind field correction.

10. A flood control drone scheduling method based on digital elevation and meteorological models according to claim 1, characterized in that, When using the improved ant colony algorithm, the pheromone volatilization coefficient ρ is dynamically adjusted according to the real-time meteorological risk level: when the regional rainfall intensity is greater than 50 mm / h, the pheromone volatilization coefficient ρ is increased to 0.25.

Images