Multi-scale progressive short-term flood intelligent early warning method, device, equipment and medium

Abstract

The invention relates to a multi-scale progressive short-term flood intelligent early warning method, device and equipment and a medium, and the method comprises the steps: obtaining rainfall forecast data of a target region within a preset time period, and driving a low-resolution numerical model to carry out the numerical calculation, thereby obtaining a first-resolution flood flow field prediction result, performing partition parallel computing by using a preset model to obtain a second-resolution flood flow field prediction result, enhancing the local flood flow field resolution by using another preset model to obtain a third-resolution flood flow field prediction result, and obtaining a third-resolution flood flow field prediction result on the basis of a pixel range. And determining a current early warning level of the target area according to the maximum water depth information and the maximum flow velocity information, so as to carry out flood early warning according to the current early warning level. Therefore, the problems of low flood early warning calculation efficiency, insufficient foreseeability and data fusion and the like are solved, and a progressive system of local refined flood early warning, rapid trend analysis, accurate disaster reminding and directional early warning release is realized for dynamic scene requirements of flood.

Description

Technical Field

[0001] The present application relates to the technical field of flood disaster prediction, and in particular to a multi-scale progressive short-term flood intelligent early warning method, device, equipment and medium. Background Art

[0002] Urban stormwater flooding poses a global threat to people and infrastructure, a threat that is expected to intensify as floods become more frequent and the world's population grows, with 68% of the world's population expected to live in urban areas by 2050. From an economic perspective, the concentration of wealth in urban areas and the accelerated development of infrastructure have led to a significant increase in economic losses from flooding, and these losses are expected to increase further globally.

[0003] The flood warning system in related technologies has the following technical bottlenecks: (1) Low computational efficiency of refined models: Flood warning systems in related technologies mainly rely on hydrological and hydrodynamic coupling numerical models. When finer grid cells are used for numerical solutions, the solution time increases dramatically due to the increase in the number of grids and the decrease in the allowable time step. This creates a trade-off between simulation resolution and computing resources, making it difficult for related models to be effectively expanded and applied to large-scale flood warning applications. Therefore, how to improve the computational efficiency of flood models and effectively balance the relationship between computational accuracy and computational efficiency is a hot demand for achieving smart disaster prevention and mitigation.

[0004] (2) Insufficient predictability of intelligent models: To improve the computational efficiency of flood models, early warning systems in related technologies often use intelligent models based on deep neural networks for rapid prediction. However, due to the constraints of video memory usage, deep learning models in related technologies often adopt an autoregressive paradigm (i.e., using the flood and inundation distribution of several historical steps and the current step rainfall to predict the flood and inundation distribution of the current step) for learning. As a result, during the rolling forecast process, errors accumulate due to the increase in the number of iterations, making the predicted flood and inundation distribution fuzzy and dissipated. It is impossible to accurately analyze the short-term (0-24 hours) flood and inundation distribution at a high spatiotemporal resolution (the spatial resolution of flood and inundation is not less than 10 minutes and the spatial resolution is not less than 5 meters), making it difficult for the prediction results of deep learning models in related technologies to provide forward-looking guidance for flood control work.

[0005] (3) Insufficient ability to integrate and apply multi-source and multi-scale elevation data: Surface elevation is a key factor in controlling the evolution of floods. Considering the accuracy requirements of flood warnings, flood warning systems in related technologies often use sub-meter elevation data measured by LiDAR to construct flood models. However, due to the difficulty in applying for LiDAR measurement permissions and the high cost of measurement, related technologies are unable to achieve full coverage of sub-meter elevation data. This makes the integration and application of multi-scale elevation data obtained by multiple types of measurement methods an inevitable trend in the development of flood modeling technology. However, flood modeling in related technologies mainly uses a single-resolution regular grid, which cannot fully utilize the efficiency advantages of multi-scale modeling. As a result, existing flood warning systems often face a trade-off between the ability to perceive the overall flood evolution situation and the accuracy of analyzing the details of local flood disasters. They are unable to coordinate and achieve a coordinated early warning of cross-scale flood disaster events in megacities. Summary of the Invention

[0006] The present application provides a multi-scale progressive short-term flood intelligent early warning method, device, equipment and medium to solve the problems of low computational efficiency, insufficient foresight and insufficient data fusion application capabilities of flood early warning systems. It can effectively balance the relationship between computational efficiency and computational accuracy, and realize multi-source and multi-scale data fusion and rapid short-term flood early warning.

[0007] The first embodiment of the present application provides a multi-scale progressive short-term flood intelligent early warning method, comprising the following steps: Obtain rainfall forecast data for the target area within a preset time period; Based on the rainfall forecast data within the preset time period, driving the low-resolution numerical model to perform numerical calculations to obtain a first-resolution flood flow field prediction result; Based on the first-resolution flood flow field prediction results, the preset LSR-UNet-L1 model is used to perform partitioned parallel calculations to obtain the second-resolution flood flow field prediction results; Based on the second-resolution flood flow field prediction results, the preset LSR-UNet-L2 model is used to enhance the local flood flow field resolution to obtain the third-resolution flood flow field prediction results; The pixel range is determined based on the maximum water depth information in the third resolution flood flow field prediction result, and based on the pixel range, the current warning level of the target area is determined according to the maximum water depth information and the maximum flow velocity information in the third resolution flood flow field prediction result, so as to perform a flood warning according to the current warning level.

[0008] Optionally, in some embodiments, before performing partitioned parallel calculation using the preset LSR-UNet-L1 model based on the first-resolution flood flow field prediction result to obtain the second-resolution flood flow field prediction result, the method further includes: Determine an area to be simulated, divide the area to be simulated into multiple sub-areas, and determine constraints for each sub-area; Extracting a rainfall-flood flow field dataset for each sub-region from a preset flood flow field sequence based on the constraint conditions of each sub-region; Based on a preset loss function, a preset neural network is trained according to the rainfall-flood flow field dataset of each sub-region to obtain the preset LSR-UNet-L1 model.

[0009] Optionally, in some embodiments, before extracting the rainfall-flood flow field dataset of each sub-region from the preset flood flow field sequence based on the constraint conditions of each sub-region, the method further includes: Obtain downscaled products based on satellite precipitation datasets; Based on a preset random rainstorm displacement method, generating multiple sets of random rainfall scenarios according to the downscaled products, and generating the preset random driving dataset according to the multiple sets of random rainfall scenarios; Based on the preset fused DEM (Digital Elevation Model) data, a high-resolution flood numerical model and a low-resolution flood numerical model are constructed. The high-resolution flood numerical model is calculated based on unstructured grid discretization and hydrological and hydrodynamic coupling. Based on multiple groups of random rainfall scenarios in the preset random driving data set, numerical calculations are performed using the low-resolution flood numerical model and the high-resolution flood numerical model respectively to obtain the preset flood flow field sequence, which meets the time resolution of the preset duration.

[0010] Optionally, in some embodiments, before constructing the high-resolution flood numerical model and the low-resolution flood numerical model based on the preset fused DEM data, the method further includes: Based on the preset urban flood numerical simulation requirements, obtain raster datasets and spatial geographic datasets; Based on the raster dataset and the spatial geographic dataset, with the preset DEM-LR as a benchmark, the preset DEM-HR fusion is aligned to the preset DEM-LR to obtain the preset fused DEM data.

[0011] Optionally, in some embodiments, taking a preset DEM-LR as a reference, aligning the fusion of a preset DEM-HR to the preset DEM-LR to obtain the fused DEM data includes: Based on the measurement boundary line of the preset DEM-HR, determining the pixel mean offset between the preset DEM-HR and the preset DEM-LR along the measurement boundary line, Based on the preset DEM-HR measurement boundary line and the first preset multiple, generating a buffer zone according to the preset DEM-HR pixel size; Based on the boundary of the buffer zone, the preset DEM-HR is mosaicked into the preset DEM-LR, and based on a preset buffer smoothing strategy, the preset fused DEM data is obtained according to the mosaicking result.

[0012] Optionally, in some embodiments, the preset LSR-UNet-L2 model is: ; in, To balance the weight coefficient of flood flow field variable prediction error, is the mean square error of the flood flow field variable prediction, is the binary cross entropy error for flood inundation range prediction.

[0013] A second embodiment of the present application provides a multi-scale progressive short-term flood intelligent early warning device, comprising: An acquisition module is used to obtain rainfall forecast data for a target area within a preset time period; A first prediction module is configured to drive a low-resolution numerical model to perform numerical calculations based on the rainfall forecast data within the preset time period to obtain a first-resolution flood flow field prediction result; A second prediction module is configured to perform partitioned parallel calculations using a preset LSR-UNet-L1 model based on the first-resolution flood flow field prediction results to obtain a second-resolution flood flow field prediction result; A third prediction module is configured to enhance the local flood flow field resolution using a preset LSR-UNet-L2 model based on the second-resolution flood flow field prediction result to obtain a third-resolution flood flow field prediction result; An early warning module is used to determine a pixel range based on the maximum water depth information in the third-resolution flood flow field prediction result, and based on the pixel range, determine a current early warning level of the target area according to the maximum water depth information and the maximum flow velocity information in the third-resolution flood flow field prediction result, so as to perform a flood early warning according to the current early warning level.

[0014] Optionally, in some embodiments, before performing partitioned parallel calculation using the preset LSR-UNet-L1 model based on the first-resolution flood flow field prediction result to obtain the second-resolution flood flow field prediction result, the second prediction module further includes: A division unit, configured to determine an area to be simulated, divide the area to be simulated into a plurality of sub-areas, and determine a constraint condition for each sub-area; an extraction unit, configured to extract a rainfall-flood flow field dataset of each sub-region from a preset flood flow field sequence based on the constraint conditions of each sub-region; The generation unit is used to train a preset neural network based on a preset loss function and the rainfall-flood flow field dataset of each sub-region to obtain the preset LSR-UNet-L1 model.

[0015] Optionally, in some embodiments, before extracting the rainfall-flood flow field dataset of each sub-region from the preset flood flow field sequence based on the constraint conditions of each sub-region, the extraction unit further includes: Get subunits for obtaining downscaled products based on satellite precipitation datasets; A generating subunit, configured to generate a plurality of random rainfall scenarios according to the downscaled product based on a preset random rainstorm displacement method, and generate the preset random driving dataset according to the plurality of random rainfall scenarios; A construction subunit is used to construct a high-resolution flood numerical model and a low-resolution flood numerical model based on preset fused DEM data, wherein the high-resolution flood numerical model is obtained based on unstructured grid discretization and hydrological and hydrodynamic coupling calculation; A computing subunit is used to perform numerical calculations based on multiple groups of random rainfall scenarios in the preset random driving data set, using the low-resolution flood numerical model and the high-resolution flood numerical model to obtain the preset flood flow field sequence, where the preset flood flow field sequence meets the time resolution of the preset duration.

[0016] Optionally, in some embodiments, before constructing the high-resolution flood numerical model and the low-resolution flood numerical model based on the preset fused DEM data, the constructing subunit further includes: Obtain sub-components, used to obtain raster datasets and spatial geographic datasets based on preset urban flood numerical simulation requirements; The fusion subcomponent is used to align the fusion of the preset DEM-HR to the preset DEM-LR based on the raster dataset and the spatial geographic dataset, so as to obtain the preset fused DEM data.

[0017] Optionally, in some embodiments, the fusion component includes: a determination component for determining, based on the measurement boundary line of the preset DEM-HR, a pixel mean offset between the preset DEM-HR and the preset DEM-LR along the measurement boundary line; A first generating subcomponent is used to generate a buffer zone according to the preset DEM-HR pixel size based on the preset DEM-HR measurement boundary line and a first preset multiple; The second generation component mosaics the preset DEM-HR into the preset DEM-LR based on the boundary of the buffer zone, and obtains the preset fused DEM data according to the mosaic result based on a preset buffer smoothing strategy.

[0018] Optionally, in some embodiments, the preset LSR-UNet-L2 model is: ; in, To balance the weight coefficient of flood flow field variable prediction error, is the mean square error of the flood flow field variable prediction, is the binary cross entropy error for flood inundation range prediction.

[0019] The third aspect of the present application provides an electronic device, comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to implement the multi-scale progressive short-term flood intelligent early warning method as described in the above embodiment.

[0020] The fourth aspect of the present application provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement the multi-scale progressive short-term flood intelligent early warning method as described in the above embodiment.

[0021] Therefore, this application has at least the following beneficial effects: (1) This application constructs a hierarchical intelligent computing structure of "global partition parallel computing-limited area super-resolution computing", which effectively takes into account the cross-scale computing needs of global trend deduction and local detail analysis, allowing users to flexibly adjust the corresponding partition computing and super-resolution computing settings based on their own computing resources, data availability and targeted warning goals, and effectively balance the relationship between computing efficiency and computing accuracy.

[0022] (2) This application constructs a digital-intelligence fusion computing strategy that uses low-resolution numerical simulation to guide deep learning intelligent computing. It uses the numerical model's robust analytical ability for the overall evolution trend of the flood flow field to overcome the fuzzy dissipation problem faced by the intelligent model during multi-step iteration. On this basis, it gives full play to the intelligent model's powerful learning ability for local high-frequency nonlinear characteristics, overcomes the computational efficiency bottleneck faced by the numerical model when conducting high-resolution simulation, and realizes rapid short-term flood forecasting with a forecast period of 24 hours and a time resolution of no less than 10 minutes.

[0023] (3) This application can align and integrate multi-source and multi-scale basic data, so that the intelligent model can gradually focus on the flood warning area in a geographic coordinate system that is as consistent as possible, improve the accuracy of flood flow field analysis, and achieve large-scale progressive and refined flood risk warning covering the entire area of ​​a megacity and with a spatial resolution of no less than 5 meters in key areas.

[0024] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which: Figure 1 A schematic diagram of the principle of the LSR-UNet model provided according to one embodiment of the present application; Figure 2 This is a flow chart of a multi-scale progressive short-term flood intelligent early warning method provided according to an embodiment of the present application; Figure 3 Schematic diagram of a multi-scale progressive short-term flood intelligent early warning device according to an embodiment of the present application; Figure 4 A schematic diagram of the structure of an electronic device provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] The following describes in detail embodiments of the present application, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0027] The following describes the multi-scale progressive short-term flood intelligent early warning method, device, equipment and medium of the embodiments of the present application with reference to the accompanying drawings. In response to the problems of low computational efficiency, insufficient foresight and insufficient data fusion and application capabilities of the flood warning system mentioned in the above background technology, the present application provides a multi-scale progressive short-term flood intelligent warning method. In this method, rainfall forecast data for the target area within a preset time period is obtained, and based on the rainfall forecast data within the preset time period, a low-resolution numerical model is driven to perform numerical calculations to obtain a first-resolution flood flow field prediction result. Based on the first-resolution flood flow field prediction result, a preset LSR-UNet-L1 model is used to perform partitioned parallel calculations to obtain a second-resolution flood flow field prediction result. Based on the second-resolution flood flow field prediction result, the preset LSR-UNet-L2 model is used to enhance the local flood flow field resolution to obtain a third-resolution flood flow field prediction result. Finally, the pixel range is determined based on the maximum water depth information in the third-resolution flood flow field prediction result, and based on the pixel range, the current warning level of the target area is determined according to the maximum water depth information and the maximum flow velocity information in the third-resolution flood flow field prediction result, so as to perform a flood warning according to the current warning level. This solves the problems of low computational efficiency, insufficient foresight, and insufficient data fusion and application capabilities of the flood warning system. It can effectively balance the relationship between computational efficiency and computational accuracy, and realize multi-source and multi-scale data fusion and rapid short-term flood warning.

[0028] Before introducing the multi-scale progressive short-term flood intelligent early warning method according to the embodiment of the present application, the LSR-UNet model is first introduced.

[0029] Specifically, if Figure 1 As shown in the figure, the LSR-UNet model uses the classic UNet network in the field of computer vision as its basic structure. The UNet network is a deep learning model based on the convolutional neural network structure, consisting of an encoder, a decoder, and skip connections: (1) Encoder module: It belongs to the input branch of the UNet network and is composed of a series of DownConv units. By continuously applying 3×3 convolution and nonlinear activation functions, it gradually extracts the distribution features of the input image from local to global, and saves and encodes them as feature images of the corresponding scale.

[0030] (2) Decoder module: It belongs to the output branch of the UNet network and is composed of a series of UpConv units. It forms a high-level feature information representation by continuously applying bilinear interpolation and DoubleConv (consisting of two consecutive 3×3 convolutions and nonlinear activations) and splicing it with the feature map of the corresponding layer. Finally, the target image is output by the OutConv unit (consisting of one 1×1 convolution and nonlinear activation).

[0031] (3) Skip connection module: It is mainly used to connect the feature map information between the decoder and the encoder, thereby allowing the decoder to directly obtain low-level feature information such as local edges and textures extracted by the encoder, making it easier for the UNet model to fully integrate and utilize multi-scale features, reducing the difficulty of network training and improving its simulation accuracy in flood flow field simulation.

[0032] In actual use, the channels of the input image can be defined by flood flow field data, geographic data, etc.

[0033] Figure 2 This is a flow chart of a multi-scale progressive short-term flood intelligent early warning method provided in an embodiment of the present application.

[0034] like Figure 2 As shown, the multi-scale progressive short-term flood intelligent early warning method includes the following steps: In step S101, rainfall forecast data for a target area within a preset time period is obtained.

[0035] The preset duration may be pre-set by the user, obtained through a limited number of experiments, or obtained through a limited number of computer simulations, and is not specifically limited here.

[0036] Specifically, when carrying out multi-scale progressive early warning for short-term flood processes of 0-24 hours, the embodiment of the present application can obtain short-term rainfall forecast data of 0-24 hours for the target area.

[0037] In step S102, based on the rainfall forecast data within a preset time period, a low-resolution numerical model is driven to perform numerical calculations to obtain a first-resolution flood flow field prediction result.

[0038] The first resolution may be preset by the user, obtained through a limited number of experiments, or obtained through a limited number of computer simulations, and is not specifically limited here. Preferably, the first resolution is 300 meters.

[0039] For example, the embodiment of the present application obtains 0-24 hours short-term rainfall forecast data, thereby driving the low-resolution numerical model Numerical calculations were performed and resampling was performed to generate flood flow field prediction results with a resolution of 300 meters.

[0040] In step S103, based on the flood flow field prediction result of the first resolution, the preset LSR-UNet-L1 model is used to perform partitioned parallel calculation to obtain the flood flow field prediction result of the second resolution.

[0041] The second resolution may be preset by the user, obtained through a limited number of experiments, or obtained through a limited number of computer simulations, and is not specifically limited here. Preferably, the second resolution is 30 meters.

[0042] For example, based on the flood flow field prediction results with a resolution of 300 meters, the preset LSR-UNet-L1 model is used for partitioned parallel calculation to generate the flood flow field prediction results with a resolution of 30 meters.

[0043] Furthermore, in some embodiments, before obtaining the second resolution flood flow field prediction result based on the first resolution flood flow field prediction result and performing partitioned parallel calculation using the preset LSR-UNet-L1 model, it also includes: determining the area to be simulated, and dividing the area to be simulated into multiple sub-areas, and determining the constraints of each sub-area; based on the constraints of each sub-area, extracting the rainfall-flood flow field data set of each sub-area from the preset flood flow field sequence; based on the preset loss function, training the preset neural network according to the rainfall-flood flow field data set of each sub-area to obtain the preset LSR-UNet-L1 model.

[0044] Specifically, the embodiment of the present application adopts the divide-and-conquer modeling idea, decomposing the area to be simulated into sub-regions Referring to the spatial resolution characteristics of commonly used public elevation datasets, the sub-region flood flow field resolution was set to 30 meters (corresponding to the global high-resolution numerical simulation resampling result) and 300 meters (corresponding to the global low-resolution numerical simulation resampling result). On this basis, in order to preserve the connection between sub-regions, take into account the boundary characteristics of irregular simulation areas and the needs of subsequent intelligent model training, the following constraints were imposed on the sub-region decomposition: (1) The overlapping area ratio between sub-regions is not less than 10%; (2) The sub-region covers no less than 70% of the internal area of ​​the study area; (3) The length and width of the sub-region counted by the number of pixels after resampling the high / low resolution flood flow field should be integer multiples of 8 (such as 64×64, 128×128, 640×640).

[0045] To facilitate subsequent model calculations, the 30-m resolution flood flow field is used as the basic grid, and the nearest neighbor sampling method is used to interpolate the 300-m flood flow field to each pixel in the basic grid so that the two maintain consistent length and width.

[0046] A rainfall-flood flow field dataset was extracted from each subregion to train the LSR-UNet-L1 intelligent model for parallel prediction of subregional flood flow fields. LSR-UNet-L1 takes 300-meter-resolution flow field variables (i.e., water depth, x-direction velocity, and y-direction velocity), 30-meter elevation (interpolated to the base grid using the nearest neighbor sampling method), and rainfall distribution (interpolated to the base grid using the nearest neighbor sampling method) as input. It outputs 30-meter-resolution flow field variables, which are split into training, validation, and test datasets in a 70:15:15 ratio. The specific training steps include: 1. Shared parameter training among sub-regions, that is, all sub-regions jointly construct a parameter Intelligent Model , the model needs to traverse all sub-region samples during training, thereby minimizing the loss function , where Based on intelligent calculation value and numerical simulation values The masked mean squared error for the cell-by-cell comparison (that is, the mean squared error for cells falling outside the study area is not considered).

[0047] 2. Fine-tuning of independent parameters within sub-regions, that is, after completing the shared parameter training, freezing the shared intelligent model in each sub-region The remaining parameters except the last output layer are fine-tuned using the corresponding sub-region samples, and finally form The parameters are Intelligent Model .

[0048] On this basis, for the connection relationship between the simulation values ​​of different sub-regions, the mean square error loss function On the basis of , the connection loss function between the sub-region and its adjacent regions is added, and it is agreed that for the units with spatial overlap between sub-regions, the average value of the intelligent model of each sub-region is used as the simulation reference value: and For example, the connection loss function based on the sample traversal of the overlapping area and It can be defined as: ; in, is the first loss function, is the second loss function, For sub-region The mean square error of the unit simulation value in the overlapping area, For sub-region Mean square error of the unit simulation values ​​in the overlapping area.

[0049] By minimizing the connection loss function and , which can guide the intelligent models of each sub-region to form variable values ​​that are as consistent as possible at the connection boundaries.

[0050] Optionally, in some embodiments, before extracting the rainfall-flood flow field dataset of each sub-region from a preset flood flow field sequence based on the constraints of each sub-region, it also includes: obtaining a downscaling product based on a satellite precipitation dataset; based on a preset random rainstorm displacement method, generating multiple groups of random rainfall scenarios according to the downscaling product, and generating a preset random driving dataset based on the multiple groups of random rainfall scenarios; based on the preset fused DEM data, constructing a high-resolution flood numerical model and a low-resolution flood numerical model, wherein the high-resolution flood numerical model is obtained based on unstructured grid discretization and hydrological and hydrodynamic coupling calculation; based on the multiple groups of random rainfall scenarios in the preset random driving dataset, numerical calculations are performed using the low-resolution flood numerical model and the high-resolution flood numerical model respectively to obtain a preset flood flow field sequence, and the preset flood flow field sequence meets the time resolution of the preset duration.

[0051] Specifically, the embodiment of the present application can generate random rainfall scenarios. In order to ensure that the deep learning model can effectively solve the flood process under rainfall-driven scenarios with different spatiotemporal characteristics, downscaled products based on satellite precipitation datasets (such as the CMORPH dataset released by the U.S. Climate Prediction Center) are collected (with a spatiotemporal resolution of not less than 1 hour and 1 kilometer, respectively, and a total number of rainfall times of not less than 100). Based on this, a preset random rainstorm displacement method is used to generate no less than 1,000 sets of random rainfall scenarios, providing a random driven dataset for subsequent flood numerical simulations.

[0052] On this basis, a no-rainfall event and a corresponding no-flooding map were introduced to guide the LSR-UNet model to correctly predict the surface water depth and flow velocity distribution under the no-rainfall scenario (i.e., zero water depth and zero flow velocity), avoiding non-physical prediction results as much as possible.

[0053] Here, we explain how to construct a high-resolution numerical model.

[0054] For the entire study area, a high-resolution flood numerical model based on unstructured grid discretization and hydrological and hydrodynamic coupling calculation is constructed. In the process of discretization of unstructured grids, the river network boundary lines and road network boundary lines are used as discrete control lines, so that the unstructured grid can be generated in accordance with the corresponding control lines, thereby ensuring that the subsequent flood numerical calculation process can fully represent the corresponding confluence boundary information. On this basis, for key areas covered by high-resolution elevation data, the maximum grid area constraint used for unstructured grid discretization is appropriately reduced, so that the grid resolution of the corresponding area can reach the meter level, thereby ensuring that the numerical calculation results of relevant key areas can provide sufficiently accurate flood flow field information to support the needs of relevant flood prevention and emergency assessment.

[0055] Here, we explain how to construct a low-resolution numerical model.

[0056] For the entire study area, referring to the aforementioned high-resolution numerical model construction method, the control constraints based on the river network boundary lines and road network boundary lines are retained, but the maximum area constraint used for unstructured grid discretization is appropriately increased, so that: (1) The total number of grids is to ; (2) The minimum grid area is not less than 10 .

[0057] This results in a corresponding low-resolution flood numerical model .

[0058] For random rainfall scenarios, high-resolution numerical models are used. and low-resolution numerical models Numerical calculations were performed to obtain a flood flow field sequence with a time resolution of 10 minutes. Specifically, the sequence consists of flood flow fields updated every 10 minutes, where the flow field information is expressed as water depth, x-direction flow velocity, and y-direction flow velocity variables stored in an unstructured grid.

[0059] After completing the flood numerical calculation, the corresponding rainfall scenarios, high-resolution model calculation results, and low-resolution model calculation results are paired and compiled into an ordered flood flow field dataset, which is used to train the preset neural network to obtain the LSR-UNet model.

[0060] To facilitate analysis, the flow field variables stored in the unstructured grid can be resampled to a regular grid with a fixed resolution (such as 30 meters, 5 meters, 2 meters, etc.) for a certain target area. The value of the regular grid variable can be defined as the area-weighted average sum of the unstructured grid variables that intersect with the regular grid.

[0061] Optionally, in some embodiments, before constructing a high-resolution flood numerical model and a low-resolution flood numerical model based on the preset fused DEM data, it also includes: obtaining a raster dataset and a spatial geographic dataset based on the preset urban flood numerical simulation requirements; based on the raster dataset and the spatial geographic dataset, with the preset DEM-LR as the benchmark, aligning the fusion of the preset DEM-HR to the preset DEM-LR to obtain the preset fused DEM data.

[0062] Specifically, referring to the requirements for urban flood numerical simulation, spatial geographic datasets such as digital elevation models (DEMs), land use types, and building area density were collected. Multi-source, multi-scale DEMs covering different areas, acquired by different measurement methods (e.g., lidar and satellite), were used as the benchmark. The remaining DEMs with smaller coverage and higher spatial resolution (abbreviated as DEM-HR) were then fused and aligned to the DEM-LR.

[0063] Optionally, in some embodiments, based on the preset DEM-LR, the fusion of the preset DEM-HR is aligned to the preset DEM-LR to obtain fused DEM data, including: based on the measurement boundary line of the preset DEM-HR, determining the pixel mean offset of the preset DEM-HR and the preset DEM-LR along the measurement boundary line, based on the measurement boundary line of the preset DEM-HR and the first preset multiple, generating a buffer according to the preset DEM-HR pixel size; based on the boundary of the buffer zone, mosaicking the preset DEM-HR to the preset DEM-LR, and based on the preset buffer smoothing strategy, obtaining the preset fused DEM data according to the mosaicking result.

[0064] Specifically, the embodiment of the present application takes the preset DEM-LR as a reference through the following steps, aligns the fusion of the preset DEM-HR to the preset DEM-LR, and obtains fused DEM data.

[0065] (1) Systematic error correction: Measure the boundary line along the DEM-HR, statistically analyze the pixel mean offset between the DEM-HR and DEM-LR along the boundary line, and deduct the mean offset from the pixel value in the DEM-HR.

[0066] (2) Boundary buffer smoothing: Measure the boundary line along the DEM-HR and generate a buffer zone 10 times the size of the DEM-HR pixel. Then, according to the boundary of the buffer zone, mosaic the DEM-HR into the DEM-LR. Then, along the boundary line, use a two-dimensional Gaussian filter with a kernel size of 5 times the size of the DEM-HR pixel to smooth the elevation data in the buffer zone. In this way, the fused DEM data DEM-LR-HR covering the range of the DEM-HR is obtained.

[0067] When constructing the flood numerical model subsequently, for the model units that fall within the DEM-HR range, the resolution of the DEM data covering the area is used as the priority order, so that the units first read the DEM-LR-HR data from the highest spatial resolution.

[0068] In step S104, based on the second-resolution flood flow field prediction result, the preset LSR-UNet-L2 model is used to enhance the local flood flow field resolution to obtain a third-resolution flood flow field prediction result.

[0069] The third resolution may be preset by the user, obtained through a limited number of experiments, or obtained through a limited number of computer simulations, and is not specifically limited here. Preferably, the third resolution is 3 meters.

[0070] For example, based on the 30-meter resolution flood flow field prediction results, focusing on key areas, the preset LSR-UNet-L2 model is used to enhance the local flood flow field resolution and generate a 3-meter resolution flood flow field prediction result.

[0071] Specifically, for key local areas with high-resolution basic information, referring to the commonly used refined flood simulation resolution, the high-resolution numerical simulation results in this area were resampled to 3 meters and 30 meters, respectively, corresponding to the local high-resolution flood flow field simulation results. (Among them, serial number Represents different flow field variable types, i.e., water depth, x-direction velocity, y-direction velocity) and local low-resolution flood flow field simulation results , and ensure that the length and width of the sub-region based on the resampled pixel count are both integer multiples of 8 (such as 64×64, 128×128, 640×640). If the initial region does not meet this dimensional condition, the region size is appropriately adjusted until it meets this dimensional condition.

[0072] by Based on the basic grid, the nearest neighbor sampling method is used to interpolate the spatial geographic data sets such as surface elevation, roughness (converted by land use type table), imperviousness (converted by land use type table), building area ratio to each pixel in the basic grid, and obtain the same High-resolution basement data with consistent length and width At the same time, in order to make LSR-UNet-L2 reuse the model structure of LSR-UNet-L1 as much as possible, refer to the approach in step S103 and use the nearest neighbor sampling method to Interpolate to the base grid so that and Maintain consistent length and width.

[0073] Here, we explain how to train the preset LSR-UNet-L2.

[0074] Specifically, high and low resolution flood flow field datasets were extracted in key local areas, and the LSR-UNet-L2 intelligent model for local flood flow field resolution enhancement was trained. Specifically, the LSR-UNet-L2 intelligent model was trained with 30-meter resolution flow field variables. and 3-meter resolution basement data As input, output 3-meter resolution flow field variables , and split it into training, validation, and test data sets in a ratio of 70:15:15, which are used for model parameter learning, optimal parameter selection, and resolution performance evaluation, respectively.

[0075] In view of the high spatial variability of meter-level resolution flood flow fields due to local micro-topography, an additional output branch for flood range prediction is introduced in the output part of the LSR-UNet-L2 intelligent model. (where is the weight coefficient assigned to different variables), a binary cross entropy error is introduced for flood range prediction to measure the accuracy of capturing flood inundation range:

[0076] in, is the mathematical expectation, is the pixel flooding probability predicted by the preset LSR-UNet-L2 intelligent model, Indicates the pixel submergence label value based on high-resolution simulated water depth classification, that is, when the simulated water depth exceeds 0.15 meters, , otherwise . The error function required for super-resolution model training can be expressed as:

[0077] Based on the above error function representation, the following two-stage training is carried out: 1. Initial training based on adaptive weight parameters: that is, using a multi-task learning algorithm, the original error function is expanded to:

[0078] in, To balance the weight coefficient of the prediction error of flood flow field variables (i.e. water depth, x-direction velocity, y-direction velocity), is the mean square error of the flood flow field variable prediction, is the binary cross entropy error for flood inundation range prediction.

[0079] This will jointly train the LSR-UNet-L2 intelligent model parameters and weight coefficients (to ensure training stability, you can further replace ), and get the optimal weight coefficient combination .

[0080] 2. Enhanced training based on fixed weight parameters: Based on the initial training results, the optimal weight coefficient and the original error function are used for enhanced training:

[0081] Thus, the final parameters of the LSR-UNet-L2 intelligent model are obtained.

[0082] In step S105, the pixel range is determined based on the maximum water depth information in the third resolution flood flow field prediction result, and based on the pixel range, the current warning level of the target area is determined according to the maximum water depth information and the maximum flow velocity information in the third resolution flood flow field prediction result, so as to perform flood warning according to the current warning level.

[0083] Specifically, the maximum water depth contained in the flood flow field with a resolution of 3 meters , maximum flow rate Information, filter the maximum water depth of the whole Pixel range, and within this range, press lie in , , , Warning levels of low risk, medium risk, relatively high risk and high risk are assigned respectively, and flood warnings are issued based on the current warning levels.

[0084] Therefore, the embodiments of this application, based on a limited-area super-resolution UNet network, utilize an encoder-decoder architecture and dynamic weight allocation to achieve real-time inference of urban flood evolution with spatiotemporal resolutions of minutes and meters, enabling localized, targeted, and precise early warnings. By integrating numerical simulation, deep learning, partitioned parallelism, and super-resolution analysis into a multi-scale, progressive intelligent computing architecture, this architecture effectively balances computational efficiency and accuracy, providing new impetus for high-performance urban disaster prevention and mitigation.

[0085] According to the multi-scale progressive short-term flood intelligent early warning method proposed in the embodiment of the present application, rainfall forecast data for the target area within a preset time period is obtained, and based on the rainfall forecast data within the preset time period, a low-resolution numerical model is driven to perform numerical calculations to obtain a first-resolution flood flow field prediction result, and based on the first-resolution flood flow field prediction result, a preset LSR-UNet-L1 model is used to perform partitioned parallel calculations to obtain a second-resolution flood flow field prediction result, and then based on the second-resolution flood flow field prediction result, the preset LSR-UNet-L2 model is used to enhance the local flood flow field resolution to obtain a third-resolution flood flow field prediction result, and finally, based on the maximum water depth information in the third-resolution flood flow field prediction result, the pixel range is determined, and based on the pixel range, the current warning level of the target area is determined according to the maximum water depth information and the maximum flow velocity information in the third-resolution flood flow field prediction result, so as to perform a flood warning according to the current warning level. This solves the problems of low computational efficiency, insufficient foresight, and insufficient data fusion and application capabilities of the flood warning system. It can effectively balance the relationship between computational efficiency and computational accuracy, and realize multi-source and multi-scale data fusion and rapid short-term flood warning.

[0086] Next, a multi-scale progressive short-term flood intelligent early warning device proposed according to an embodiment of the present application will be described with reference to the accompanying drawings.

[0087] Figure 3 It is a block diagram of a multi-scale progressive short-term flood intelligent early warning device according to an embodiment of the present application.

[0088] like Figure 3 As shown, the multi-scale progressive short-term flood intelligent early warning device 10 includes: an acquisition module 100 , a first prediction module 200 , a second prediction module 300 , a third prediction module 400 and an early warning module 500 .

[0089] The acquisition module 100 is used to obtain rainfall forecast data for a target area within a preset time period.

[0090] The first prediction module 200 is used to drive the low-resolution numerical model to perform numerical calculations based on rainfall forecast data within a preset time period to obtain a first-resolution flood flow field prediction result.

[0091] The second prediction module 300 is used to perform partitioned parallel calculations using a preset LSR-UNet-L1 model based on the first-resolution flood flow field prediction result to obtain a second-resolution flood flow field prediction result.

[0092] The third prediction module 400 is used to enhance the local flood flow field resolution based on the second resolution flood flow field prediction result using the preset LSR-UNet-L2 model to obtain a third resolution flood flow field prediction result.

[0093] The early warning module 500 is used to determine the pixel range based on the maximum water depth information in the third resolution flood flow field prediction result, and based on the pixel range, determine the current warning level of the target area according to the maximum water depth information and the maximum flow velocity information in the third resolution flood flow field prediction result, so as to perform flood warning according to the current warning level.

[0094] Optionally, in some embodiments, before obtaining the second resolution flood flow field prediction result based on the first resolution flood flow field prediction result by performing partitioned parallel calculation using the preset LSR-UNet-L1 model, the second prediction module 300 further includes: a division unit, an extraction unit, and a generation unit.

[0095] The division unit is used to determine the area to be simulated, divide the area to be simulated into multiple sub-areas, and determine the constraint conditions of each sub-area.

[0096] The extraction unit is used to extract the rainfall-flood flow field data set of each sub-region from the preset flood flow field sequence based on the constraint conditions of each sub-region.

[0097] The generation unit is used to train a preset neural network based on a preset loss function and a rainfall-flood flow field dataset of each sub-region to obtain a preset LSR-UNet-L1 model.

[0098] Optionally, in some embodiments, before extracting the rainfall-flood flow field dataset of each sub-area from a preset flood flow field sequence based on the constraints of each sub-area, the extraction unit further includes: an acquisition sub-unit, a generation sub-unit, a construction sub-unit and a calculation sub-unit.

[0099] Among them, the acquisition subunit is used to obtain downscaled products based on satellite precipitation datasets.

[0100] The generation subunit is used to generate multiple sets of random rainfall scenarios according to the downscaled products based on the preset random rainstorm displacement method, and to generate preset random driving data sets according to the multiple sets of random rainfall scenarios.

[0101] Construct subunits for constructing high-resolution flood numerical models and low-resolution flood numerical models based on preset fused DEM data. The high-resolution flood numerical model is calculated based on unstructured grid discretization and hydrological and hydrodynamic coupling.

[0102] The computing subunit is used to perform numerical calculations based on multiple sets of random rainfall scenarios in a preset random driving data set using a low-resolution flood numerical model and a high-resolution flood numerical model, respectively, to obtain a preset flood flow field sequence, which meets the time resolution of the preset duration.

[0103] Optionally, in some embodiments, before constructing the high-resolution flood numerical model and the low-resolution flood numerical model based on the preset fused DEM data, constructing the subunit further includes: acquiring subcomponents and fusing subcomponents.

[0104] Among them, the acquisition component is used to obtain raster datasets and spatial geographic datasets based on preset urban flood numerical simulation requirements.

[0105] The fusion component is used to align the fusion of the preset DEM-HR to the preset DEM-LR based on the raster dataset and the spatial geographic dataset, taking the preset DEM-LR as the benchmark, to obtain the preset fused DEM data.

[0106] Optionally, in some embodiments, the fusion sub-component includes: a determination sub-component, a first generation sub-component, and a second generation sub-component.

[0107] The determination subcomponent is used to determine the pixel mean offset between the preset DEM-HR and the preset DEM-LR along the measurement boundary line based on the measurement boundary line of the preset DEM-HR.

[0108] The first generating component is used to generate a buffer zone according to a preset DEM-HR pixel size based on a preset DEM-HR measurement boundary line and a first preset multiple.

[0109] The second generation component is used to mosaic the preset DEM-HR into the preset DEM-LR based on the boundary of the buffer zone, and obtain the preset fused DEM data according to the mosaic result based on the preset buffer smoothing strategy.

[0110] Optionally, in some embodiments, the preset LSR-UNet-L2 model is: ; in, To balance the weight coefficient of the prediction error of flood flow field variables (i.e. water depth, x-direction velocity, y-direction velocity), is the mean square error of the flood flow field variable prediction, is the binary cross entropy error for flood inundation range prediction.

[0111] It should be noted that the above explanation of the embodiment of the multi-scale progressive short-term flood intelligent early warning method is also applicable to the multi-scale progressive short-term flood intelligent early warning device of this embodiment, and will not be repeated here.

[0112] According to the multi-scale progressive short-term flood intelligent early warning device proposed in the embodiment of the present application, rainfall forecast data for the target area within a preset time period is obtained. Based on the rainfall forecast data within the preset time period, a low-resolution numerical model is driven to perform numerical calculations to obtain a first-resolution flood flow field prediction result. Based on the first-resolution flood flow field prediction result, a preset LSR-UNet-L1 model is used to perform partitioned parallel calculations to obtain a second-resolution flood flow field prediction result. Based on the second-resolution flood flow field prediction result, the preset LSR-UNet-L2 model is used to enhance the local flood flow field resolution to obtain a third-resolution flood flow field prediction result. Finally, based on the maximum water depth information in the third-resolution flood flow field prediction result, the pixel range is determined. Based on the pixel range, the current warning level of the target area is determined according to the maximum water depth information and the maximum flow velocity information in the third-resolution flood flow field prediction result, so as to perform a flood warning according to the current warning level. This solves the problems of low computational efficiency, insufficient foresight, and insufficient data fusion and application capabilities of the flood warning system. It can effectively balance the relationship between computational efficiency and computational accuracy, and realize multi-source and multi-scale data fusion and rapid short-term flood warning.

[0113] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device may include: Memory 401 , processor 402 , and computer programs stored in the memory 401 and executable on the processor 402 .

[0114] When the processor 402 executes the program, the multi-scale progressive short-term flood intelligent early warning method provided in the above embodiment is implemented.

[0115] Furthermore, the electronic device further includes: The communication interface 403 is used for communication between the memory 401 and the processor 402 .

[0116] The memory 401 is used to store computer programs that can be run on the processor 402 .

[0117] The memory 401 may include a high-speed RAM (Random Access Memory) memory, and may also include a non-volatile memory, such as at least one disk memory.

[0118] If the memory 401, processor 402, and communication interface 403 are implemented independently, the communication interface 403, memory 401, and processor 402 can be connected to each other via a bus and communicate with each other. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0119] Optionally, in a specific implementation, if the memory 401, the processor 402 and the communication interface 403 are integrated on a chip, the memory 401, the processor 402 and the communication interface 403 can communicate with each other through an internal interface.

[0120] The processor 402 may be a CPU (Central Processing Unit), or an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application.

[0121] An embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned multi-scale progressive short-term flood intelligent early warning method.

[0122] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0123] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of technical features indicated. Thus, a feature specified as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of this application, "N" means at least two, for example, two, three, etc., unless otherwise specifically defined.

[0124] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application belong.

[0125] It should be understood that various parts of the present application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiment, the N steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array, a field programmable gate array, etc.

[0126] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0127] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A multi-scale progressive short-term flood intelligent early warning method, characterized by: The following steps are involved: Obtain rainfall forecast data for the target area within a preset time period; Based on the rainfall forecast data within the preset time period, driving the low-resolution numerical model to perform numerical calculations to obtain a first-resolution flood flow field prediction result; Based on the first-resolution flood flow field prediction results, the preset LSR-UNet-L1 model is used to perform partitioned parallel calculations to obtain the second-resolution flood flow field prediction results; Based on the second-resolution flood flow field prediction results, the preset LSR-UNet-L2 model is used to enhance the local flood flow field resolution to obtain the third-resolution flood flow field prediction results; The pixel range is determined based on the maximum water depth information in the third resolution flood flow field prediction result, and based on the pixel range, the current warning level of the target area is determined according to the maximum water depth information and the maximum flow velocity information in the third resolution flood flow field prediction result, so as to perform a flood warning according to the current warning level.

2. The method according to claim 1, characterized in that Before obtaining the second-resolution flood flow field prediction result by performing partitioned parallel calculations using the preset LSR-UNet-L1 model based on the first-resolution flood flow field prediction result, the method further includes: Determine an area to be simulated, divide the area to be simulated into multiple sub-areas, and determine constraints for each sub-area; Extracting a rainfall-flood flow field dataset for each sub-region from a preset flood flow field sequence based on the constraint conditions of each sub-region; Based on a preset loss function, a preset neural network is trained according to the rainfall-flood flow field dataset of each sub-region to obtain the preset LSR-UNet-L1 model.

3. The method according to claim 2, characterized in that Before extracting the rainfall-flood flow field dataset of each sub-region from the preset flood flow field sequence based on the constraint conditions of each sub-region, the method further includes: Obtain downscaled products based on satellite precipitation datasets; Based on a preset random rainstorm displacement method, generating multiple sets of random rainfall scenarios according to the downscaled products, and generating the preset random driving dataset according to the multiple sets of random rainfall scenarios; Based on the preset fused digital elevation model (DEM) data, a high-resolution flood numerical model and a low-resolution flood numerical model are constructed, wherein the high-resolution flood numerical model is obtained based on unstructured grid discretization and hydrological and hydrodynamic coupling calculation; Based on multiple groups of random rainfall scenarios in the preset random driving data set, numerical calculations are performed using the low-resolution flood numerical model and the high-resolution flood numerical model respectively to obtain the preset flood flow field sequence, which meets the time resolution of the preset duration.

4. The method according to claim 3, characterized in that Before constructing the high-resolution flood numerical model and the low-resolution flood numerical model based on the preset fused DEM data, the method further includes: Based on the preset urban flood numerical simulation requirements, obtain raster datasets and spatial geographic datasets; Based on the raster dataset and the spatial geographic dataset, with the preset DEM-LR as a benchmark, the preset DEM-HR fusion is aligned to the preset DEM-LR to obtain the preset fused DEM data.

5. The method according to claim 4, characterized in that The method of aligning the fusion of the preset DEM-HR to the preset DEM-LR based on the preset DEM-LR to obtain the fused DEM data includes: Based on the measurement boundary line of the preset DEM-HR, determining the pixel mean offset between the preset DEM-HR and the preset DEM-LR along the measurement boundary line, Based on the preset DEM-HR measurement boundary line and the first preset multiple, generating a buffer zone according to the preset DEM-HR pixel size; Based on the boundary of the buffer zone, the preset DEM-HR is mosaicked into the preset DEM-LR, and based on a preset buffer smoothing strategy, the preset fused DEM data is obtained according to the mosaicking result.

6. The method according to any one of claims 1 to 5, characterized in that The preset LSR-UNet-L2 model is: ； in, To balance the weight coefficient of flood flow field variable prediction error, is the mean square error of the flood flow field variable prediction, is the binary cross entropy error for flood inundation range prediction.

7. A multi-scale progressive short-term flood intelligent early warning device, characterized in that: include: An acquisition module is used to obtain rainfall forecast data for a target area within a preset time period; A first prediction module is configured to drive a low-resolution numerical model to perform numerical calculations based on the rainfall forecast data within the preset time period to obtain a first-resolution flood flow field prediction result; A second prediction module is configured to perform partitioned parallel calculations using a preset LSR-UNet-L1 model based on the first-resolution flood flow field prediction result to obtain a second-resolution flood flow field prediction result; A third prediction module is configured to enhance the local flood flow field resolution using a preset LSR-UNet-L2 model based on the second-resolution flood flow field prediction result to obtain a third-resolution flood flow field prediction result; An early warning module is used to determine a pixel range based on the maximum water depth information in the third-resolution flood flow field prediction result, and based on the pixel range, determine a current early warning level of the target area according to the maximum water depth information and the maximum flow velocity information in the third-resolution flood flow field prediction result, so as to perform a flood early warning according to the current early warning level.

8. The device according to claim 7, characterized in that Before obtaining the second-resolution flood flow field prediction result by performing partitioned parallel calculation using the preset LSR-UNet-L1 model based on the first-resolution flood flow field prediction result, the second prediction module further includes: A division unit, configured to determine an area to be simulated, divide the area to be simulated into a plurality of sub-areas, and determine a constraint condition for each sub-area; an extraction unit, configured to extract a rainfall-flood flow field dataset of each sub-region from a preset flood flow field sequence based on the constraint conditions of each sub-region; The generation unit is used to train a preset neural network based on a preset loss function and the rainfall-flood flow field dataset of each sub-region to obtain the preset LSR-UNet-L1 model.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the multi-scale progressive short-term flood intelligent early warning method according to any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the multi-scale progressive short-term flood intelligent early warning method as described in any one of claims 1 to 6.

Images