An urban flood warning system and method based on artificial intelligence
By constructing graph structure data and urban flood warning model, considering the mutual influence between grid areas, the problem of low prediction accuracy of water accumulation depth in the existing technology is solved, and a higher-precision urban flood warning is achieved.
Patent Information
- Application Number
- CN202411233402.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-04
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-09-04
AI Technical Summary
The existing urban flood warning system only considers the data changes in the time dimension of a single grid area, and fails to effectively consider the mutual influence relationship between different grid areas, resulting in a low prediction accuracy of water accumulation depth.
By dividing urban areas into multiple grid areas, meteorological, hydrological and geographical data are collected to generate feature sequences, and graph structure data is constructed to represent the mutual influence between grid areas. The urban flood warning model is used to predict the future of each grid area. The depth of water accumulation in the time period.
The prediction accuracy of the water accumulation depth in the grid area is improved, and the time dimension changes of all grid areas are taken into account through the information aggregation of graph structure data to generate a more accurate flood warning.
Smart Images

Figure CN118735312B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of urban flood warning technology, and more specifically, to an urban flood warning system and method based on artificial intelligence. Background Art
[0002] With the impact of global climate change, the accelerated urbanization process and the aging of urban drainage systems, the risk of urban flood disasters has increased. Therefore, establishing an efficient urban flood early warning system is crucial to protecting the lives and property of urban residents.
[0003] The existing urban flood warning system grids the urban area, collects the meteorological data and hydrological data of each grid area to construct a time series, and analyzes the time series of each grid area through a time series analysis model to predict the depth of waterlogging. The time series analysis model can be an LSTM (long short-term memory neural network model) or a Transformer (converter model), etc. In addition, the existing urban flood warning system also uses a target detection model to identify the type of water splashes when vehicles pass through the waterlogged area and the number of times the type of water splashes appears to determine the depth of waterlogging in the waterlogged area. The target detection model can be a YOLO series model, etc.
[0004] However, the above scheme only considers the data changes of a single grid area in the time dimension, and does not take into account the mutual influence relationship between different grid areas. For example, the terrain difference between different grid areas will affect the direction of water flow. The water in the grid area with higher terrain will flow to the grid area with lower terrain. Therefore, only analyzing the data changes of a single grid area in the time dimension may lead to low prediction accuracy of the water depth in the grid area. Summary of the invention
[0005] The present invention provides an urban flood warning system and method based on artificial intelligence to solve the technical problems in the above-mentioned background technology.
[0006] The present invention provides an urban flood warning system based on artificial intelligence, comprising:
[0007] A grid area division module, which is used to divide the urban area into M grid areas through rectangular frames;
[0008] A feature sequence generation module, which is used to collect meteorological data, hydrological data and geographical data of M grid areas at a preset time interval r within a first preset time period R, and generate a feature sequence for each grid area;
[0009] Meteorological data include: rainfall, temperature, wind speed, wind direction and air pressure;
[0010] Hydrological data include: river flow, groundwater level, surface runoff and evaporation;
[0011] Geographic data include: elevation, slope, and landform type;
[0012] The characteristic sequence includes N sequence units, and the nth sequence unit represents the meteorological data, hydrological data, and geographic data of the grid area at the nth time point, where 1≤n≤N, N=R / r;
[0013] A graph structure data construction module, which is used to construct graph structure data according to the feature sequences of M grid areas;
[0014] The graph structure data includes: M nodes, initial features of nodes, and edges between nodes;
[0015] The M nodes establish mapping relationships with the M grid areas respectively;
[0016] The initial features of the nodes are represented by the feature sequence of the grid area that establishes a mapping relationship with the nodes;
[0017] An urban flood warning module, which is used to input the graph structure data into the urban flood warning model and output the water depth of M grid areas within the second preset time period G;
[0018] The flood control measures generation module is used to generate flood control measures according to the water accumulation depth of the M grid areas within the second preset time period G.
[0019] Furthermore, the size of the rectangular box, M, R, r, and G are all custom parameters.
[0020] Furthermore, wind direction is represented by an angle of 0° to 360°, where 0° or 360° represents north wind, 90° represents east wind, 180° represents south wind, and 270° represents west wind; slope is represented by the ground height difference divided by the horizontal distance on the surface; landform type is represented by the urbanized area divided by the grid area, and the urbanized area includes: road area and building area.
[0021] Furthermore, the edges between nodes include: if the difference between the altitude of the grid area that establishes a mapping relationship with the node and the altitude of the adjacent grid area is greater than or equal to a preset altitude threshold, a directed edge is constructed between the node and the adjacent node, wherein the preset altitude threshold is a custom parameter; if the difference between the slope of the grid area that establishes a mapping relationship with the node and the slope of the adjacent grid area is greater than or equal to a preset slope threshold, a directed edge is constructed between the node and the adjacent node, wherein the preset slope threshold is a custom parameter; if the wind direction of the grid area that establishes a mapping relationship with the node and the angle between the node and the adjacent node are less than or equal to a preset angle threshold, a directed edge is constructed between the node and the adjacent node, wherein the preset angle threshold is a custom parameter.
[0022] Further, the urban flood warning model includes: M first hidden layers, M first extraction layers, 1 second hidden layer, M second extraction layers and M first classifiers;
[0023] The M first hidden layers share weight parameters and bias parameters;
[0024] The first hidden layer includes N hidden units;
[0025] The nth hidden unit of the mth first hidden layer inputs the nth sequence unit of the feature sequence of the mth node of the graph structure data, and outputs the first update vector, where 1≤m≤M, 1≤n≤N;
[0026] The mth first extraction layer is used to extract the first update vector output by the Nth hidden unit of the mth first hidden layer as the initial feature of the mth node;
[0027] That is, the initial features of the node are converted from the feature sequence to the first update vector representation;
[0028] The second hidden layer inputs graph structure data and outputs an update matrix, where a row vector of the update matrix corresponds to a second update vector of a node;
[0029] The mth second extraction layer is used to extract the mth row vector of the update matrix and input it into the mth first classifier. The classification space of the mth first classifier represents the water depth of the grid area that establishes a mapping relationship with the mth node within the second preset time period G.
[0030] Furthermore, the calculation formula of the nth hidden unit of the mth first hidden layer includes:
[0031] ;
[0032] ;
[0033] ;
[0034] ;
[0035] Among them, 1≤u≤12, 1≤v≤12, represents the first update vector output by the nth hidden unit of the mth first hidden layer, The nth sequence unit of the feature sequence of the mth node of the graph structure data input by the nth hidden unit of the mth first hidden layer, express The u-th dimension value of express The upper triangular matrix of express The number of rows and columns of the lower triangular matrix, upper triangular matrix and lower triangular matrix is the same as have the same number of dimensions, The number of dimensions is 12, , , , , and Respectively represent the intermediate vector, the first weight parameter, the second weight parameter, the third weight parameter, the fourth weight parameter and the bias parameter of the nth hidden unit of the mth first hidden layer, sigmoid represents the sigmoid activation function, T represents the transposition operation, represents the first update vector output by the n-1th hidden unit of the mth first hidden layer, Assign a value of 0.
[0036] Furthermore, the calculation formula of the second hidden layer includes:
[0037] ;
[0038] ;
[0039] ;
[0040] Where Z represents the update matrix of the second hidden layer output, represents the second update vector of the mth node, represents the set of nodes that build directed edges with the mth node, and represents the first update vector of the mth node and the kth node, and Respectively represent the weight parameters corresponding to the mth node and the kth node, Represents the weight parameter corresponding to the directed edge from the kth node to the mth node, represents the number of directed edges from the kth node to the mth node, e represents a natural constant, sigmoid represents the sigmoid activation function, It represents a stacking operation on the second update vectors of the M nodes.
[0041] Furthermore, the historical water accumulation depth is used as the sample label of the training sample of the urban flood warning model. Before training the urban flood warning model, the first hidden layer is pre-trained, and the first update vector output by the Nth hidden unit of the first hidden layer is input to the second classifier, and the classification space of the second classifier represents the rainfall at the next time point.
[0042] Furthermore, four warning levels are divided according to the depth of water accumulation in each grid area within the second preset time period G. The first-level warning corresponds to a water accumulation depth between 0 and 10 cm, and the corresponding flood prevention and control measures are to issue warning information to remind citizens to pay attention to weather changes. The second-level warning corresponds to a water accumulation depth between 10 cm and 30 cm, and the corresponding flood prevention and control measures are to strengthen monitoring and patrols, and report water accumulation in a timely manner. The third-level warning corresponds to a water accumulation depth between 30 cm and 50 cm, and the corresponding flood prevention and control measures are to remind citizens to pay attention to safety and avoid going to waterlogged areas. The fourth-level warning corresponds to a water accumulation depth of more than 50 cm, and the corresponding flood prevention and control measures are to issue emergency notices to remind citizens to stay away from waterlogged areas.
[0043] Furthermore, the present invention provides an urban flood warning method based on artificial intelligence, comprising the following steps:
[0044] Step S201, dividing the urban area into M grid areas by rectangular frames;
[0045] Step S202, within a first preset time period R, collecting meteorological data, hydrological data and geographical data of M grid areas at preset time intervals r, and generating a feature sequence for each grid area;
[0046] Step S203, constructing graph structure data according to the feature sequences of the M grid areas;
[0047] Step S204, inputting the graph structure data into the urban flood warning model, and outputting the water depth of the M grid areas within the second preset time period G;
[0048] Step S205: generating flood prevention and control measures according to the water depths of the M grid areas within the second preset time period G.
[0049] The beneficial effects of the present invention are as follows: the present invention represents the mutual influence relationship between grid areas by constructing directed edges of graph structure data, considers the data changes of all grid areas in the time dimension through the urban flood warning model, and aggregates information of the node data in the graph structure data, thereby improving the prediction accuracy of the water depth of the grid area. BRIEF DESCRIPTION OF THE DRAWINGS
[0050] Figure 1 is a schematic diagram of an artificial intelligence-based urban flood warning system of the present invention;
[0051] Figure 2 It is a flow chart of an urban flood warning method based on artificial intelligence of the present invention.
[0052] In the figure: a grid area division module 101, a feature sequence generation module 102, a graph structure data construction module 103, an urban flood warning module 104, and a flood prevention and control measures generation module 105. DETAILED DESCRIPTION
[0053] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.
[0054] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.
[0055] like Figure 1-2 As shown, an urban flood warning system based on artificial intelligence includes:
[0056] A grid area division module 101, which is used to divide the urban area into M grid areas through rectangular frames;
[0057] A feature sequence generating module 102 is used to collect meteorological data, hydrological data and geographical data of M grid areas at a preset time interval r within a first preset time period R, and generate a feature sequence for each grid area;
[0058] Meteorological data include: rainfall, temperature, wind speed, wind direction and air pressure;
[0059] Hydrological data include: river flow, groundwater level, surface runoff and evaporation;
[0060] Geographic data include: elevation, slope, and landform type;
[0061] The characteristic sequence includes N sequence units, and the nth sequence unit represents the meteorological data, hydrological data, and geographic data of the grid area at the nth time point, where 1≤n≤N, N=R / r;
[0062] A graph structure data construction module 103, which is used to construct graph structure data according to the feature sequences of the M grid areas;
[0063] The graph structure data includes: M nodes, initial features of nodes, and edges between nodes;
[0064] The M nodes establish mapping relationships with the M grid areas respectively;
[0065] The initial features of the nodes are represented by the feature sequence of the grid area that establishes a mapping relationship with the nodes;
[0066] The urban flood warning module 104 is used to input the graph structure data into the urban flood warning model and output the water depth of M grid areas within the second preset time period G;
[0067] The flood control measures generating module 105 is used to generate flood control measures according to the water accumulation depths of the M grid areas within the second preset time period G.
[0068] In one embodiment of the present invention, the size of the rectangular box, M, R, r and G are all custom parameters, M is evaluated and set by relevant experts according to the size of the urban area, the sizes of the matrix boxes can be the same or different, preferably, R is set to 1 hour, r is set to 5 minutes, then N=R / r=12, and G is set to 10 minutes.
[0069] In one embodiment of the present invention, rainfall, temperature, wind speed, wind direction and air pressure are collected by a rain gauge, a temperature sensor, an anemometer, a wind vane and a barometer respectively.
[0070] In one embodiment of the present invention, river flow refers to the amount of water in a river section per unit time, in cubic meters per second. Surface runoff refers to the amount of water that flows along the surface after rainfall without being absorbed by the soil. Evaporation refers to the total amount of water loss caused by plant transpiration and surface evaporation. Evaporation can be calculated using the Penman formula, which will not be elaborated here.
[0071] In one embodiment of the present invention, multiple data collection points are set in each grid area, and the maximum value or average value of the meteorological data, hydrological data and geographical data of the multiple data collection points is taken as the meteorological data, hydrological data and geographical data of one grid area.
[0072] In one embodiment of the present invention, the wind direction is represented by an angle of 0° to 360°, wherein 0° or 360° represents north wind, 90° represents east wind, 180° represents south wind, and 270° represents west wind; the slope is represented by the ground height difference divided by the horizontal distance on the surface; the landform type is represented by the urbanization area divided by the grid area, and the urbanization area includes: road area and building area.
[0073] In one embodiment of the present invention, the slope can also be expressed by an angle, and the calculation formula is as follows: .
[0074] In one embodiment of the present invention, the edges between nodes include: if the difference between the altitude of the grid area that establishes a mapping relationship with the node and the altitude of the adjacent grid area is greater than or equal to a preset altitude threshold, a directed edge is constructed between the node and the adjacent node, wherein the preset altitude threshold is a custom parameter, and preferably, the preset altitude threshold is set to 1 meter; if the difference between the slope of the grid area that establishes a mapping relationship with the node and the slope of the adjacent grid area is greater than or equal to a preset slope threshold, a directed edge is constructed between the node and the adjacent node, wherein the preset slope threshold is a custom parameter, and preferably, the preset slope threshold is set to 5°; if the wind direction of the grid area that establishes a mapping relationship with the node and the angle between the node and the adjacent node are less than or equal to the preset angle threshold, a directed edge is constructed between the node and the adjacent node, wherein the preset angle threshold is a custom parameter, and preferably, the preset angle threshold is set to 30°.
[0075] In one embodiment of the present invention, the urban flood warning model includes: M first hidden layers, M first extraction layers, 1 second hidden layer, M second extraction layers and M first classifiers;
[0076] The M first hidden layers share weight parameters and bias parameters;
[0077] The first hidden layer includes N hidden units;
[0078] The nth hidden unit of the mth first hidden layer inputs the nth sequence unit of the feature sequence of the mth node of the graph structure data, and outputs the first update vector, where 1≤m≤M, 1≤n≤N;
[0079] The mth first extraction layer is used to extract the first update vector output by the Nth hidden unit of the mth first hidden layer as the initial feature of the mth node;
[0080] That is, the initial features of the node are converted from the feature sequence to the first update vector representation;
[0081] The second hidden layer inputs graph structure data and outputs an update matrix, where a row vector of the update matrix corresponds to a second update vector of a node;
[0082] The mth second extraction layer is used to extract the mth row vector of the update matrix and input it into the mth first classifier. The classification space of the mth first classifier represents the water depth of the grid area that establishes a mapping relationship with the mth node within the second preset time period G.
[0083] In one embodiment of the present invention, the first classifier is a softmax classifier, and M first classifiers share weight parameters and bias parameters, which are not described in detail here.
[0084] In one embodiment of the present invention, the calculation formula of the nth hidden unit of the mth first hidden layer includes:
[0085] ;
[0086] ;
[0087] ;
[0088] ;
[0089] Among them, 1≤u≤12, 1≤v≤12, represents the first update vector output by the nth hidden unit of the mth first hidden layer, The nth sequence unit of the feature sequence of the mth node of the graph structure data input by the nth hidden unit of the mth first hidden layer, express The u-th dimension value of express The upper triangular matrix of express The number of rows and columns of the lower triangular matrix, upper triangular matrix and lower triangular matrix is the same as have the same number of dimensions, The number of dimensions is 12, , , , , and Respectively represent the intermediate vector, the first weight parameter, the second weight parameter, the third weight parameter, the fourth weight parameter and the bias parameter of the nth hidden unit of the mth first hidden layer, sigmoid represents the sigmoid activation function, T represents the transposition operation, represents the first update vector output by the n-1th hidden unit of the mth first hidden layer, Assign a value of 0.
[0090] It should be noted that The number of dimensions is 12, and the 12 dimension values correspond to rainfall, temperature, wind speed, wind direction, air pressure, river flow, groundwater level, surface runoff, evaporation, altitude, slope and landform type.
[0091] It should be noted that the first weight parameter, the second weight parameter, the third weight parameter, the fourth weight parameter and the bias parameter are all learnable parameters. The size after transposition is 12×1, so the size of the upper triangular matrix and the lower triangular matrix are both 12×12, so the size of the third weight parameter and the third weight parameter is 12×1, the size of the intermediate vector is also 12×1, the size of the first weight parameter can be designed to be 16×12, and the size after multiplication with the intermediate vector is 16×1, then the size of the second weight parameter is 16×16, and the size of the first update vector is 16×1.
[0092] In one embodiment of the present invention, the nth hidden unit of the mth first hidden layer may also be constructed based on an LSTM unit or a GRU (Gated Recurrent Network) unit.
[0093] In one embodiment of the present invention, the calculation formula of the second hidden layer includes:
[0094] ;
[0095] ;
[0096] ;
[0097] Where Z represents the update matrix of the second hidden layer output, represents the second update vector of the mth node, represents the set of nodes that build directed edges with the mth node, and represents the first update vector of the mth node and the kth node, and Respectively represent the weight parameters corresponding to the mth node and the kth node, Represents the weight parameter corresponding to the directed edge from the kth node to the mth node, represents the number of directed edges from the kth node to the mth node, e represents a natural constant, sigmoid represents the sigmoid activation function, It represents a stacking operation on the second update vectors of the M nodes.
[0098] It should be noted that, during the training process of the urban flood warning model, the maximum number of training iterations is set, the difference between the value output by the urban flood warning model at each iteration and the sample label of the training sample is specified as the loss function, and the chain rule is used to calculate the gradient of the loss function with respect to each weight parameter and bias parameter. According to the calculated gradient and the preset learning rate, the weight parameters and bias parameters of the urban flood warning model are updated.
[0099] In one embodiment of the present invention, the historical water accumulation depth is used as the sample label of the training sample of the urban flood warning model. Before training the urban flood warning model, the first hidden layer is pre-trained, and the first update vector output by the Nth hidden unit of the first hidden layer is input to the second classifier, and the classification space of the second classifier represents the rainfall at the next time point.
[0100] It should be noted that pre-training can improve the feature extraction capability of the first hidden layer for feature sequences, thereby accelerating the training speed of the urban flood warning model to a certain extent.
[0101] In one embodiment of the present invention, four warning levels are divided according to the depth of water accumulation in each grid area within the second preset time period G. The first-level warning corresponds to a water accumulation depth between 0 and 10 cm, and the corresponding flood prevention and control measures are to issue warning information to remind citizens to pay attention to weather changes. The second-level warning corresponds to a water accumulation depth between 10 cm and 30 cm, and the corresponding flood prevention and control measures are to strengthen monitoring and patrols, and report water accumulation in a timely manner. The third-level warning corresponds to a water accumulation depth between 30 cm and 50 cm, and the corresponding flood prevention and control measures are to remind citizens to pay attention to safety and avoid going to waterlogged areas. The fourth-level warning corresponds to a water accumulation depth of more than 50 cm, and the corresponding flood prevention and control measures are to issue emergency notices to remind citizens to stay away from waterlogged areas.
[0102] In one embodiment of the present invention, Figure 2 As shown, an urban flood warning method based on artificial intelligence includes the following steps:
[0103] Step S201, dividing the urban area into M grid areas by rectangular frames;
[0104] Step S202, within a first preset time period R, collecting meteorological data, hydrological data and geographical data of M grid areas at preset time intervals r, and generating a feature sequence for each grid area;
[0105] Step S203, constructing graph structure data according to the feature sequences of the M grid areas;
[0106] Step S204, inputting the graph structure data into the urban flood warning model, and outputting the water depth of the M grid areas within the second preset time period G;
[0107] Step S205: generating flood prevention and control measures according to the water depths of the M grid areas within the second preset time period G.
[0108] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.
Claims
1. An urban flood warning system based on artificial intelligence, characterized in that: include: A grid area division module, which is used to divide the urban area into M grid areas through rectangular frames; A feature sequence generation module, which is used to collect meteorological data, hydrological data and geographical data of M grid areas at a preset time interval r within a first preset time period R, and generate a feature sequence for each grid area; Meteorological data include: rainfall, temperature, wind speed, wind direction and air pressure; Hydrological data include: river flow, groundwater level, surface runoff and evaporation; Geographic data include: elevation, slope, and landform type; The characteristic sequence includes N sequence units, and the nth sequence unit represents the meteorological data, hydrological data, and geographic data of the grid area at the nth time point, where 1≤n≤N, N=R / r; A graph structure data construction module, which is used to construct graph structure data according to the feature sequences of M grid areas; The graph structure data includes: M nodes, initial features of nodes, and edges between nodes; The M nodes establish mapping relationships with the M grid areas respectively; The initial features of the nodes are represented by the feature sequence of the grid area that establishes a mapping relationship with the nodes; An urban flood warning module, which is used to input the graph structure data into the urban flood warning model and output the water depth of M grid areas within the second preset time period G; A flood control measures generation module, which is used to generate flood control measures according to the water depth of the M grid areas within the second preset time period G; The edges between nodes include: if the difference between the altitude of the grid area with which the node is mapped and the altitude of the adjacent grid area is greater than or equal to a preset altitude threshold, a directed edge is constructed between the node and the adjacent node, wherein the preset altitude threshold is a custom parameter; if the difference between the slope of the grid area with which the node is mapped and the slope of the adjacent grid area is greater than or equal to a preset slope threshold, a directed edge is constructed between the node and the adjacent node, wherein the preset slope threshold is a custom parameter; if the wind direction of the grid area with which the node is mapped and the angle between the node and the adjacent node are less than or equal to a preset angle threshold, a directed edge is constructed between the node and the adjacent node, wherein the preset angle threshold is a custom parameter; The urban flood warning model includes: M first hidden layers, M first extraction layers, 1 second hidden layer, M second extraction layers and M first classifiers; The M first hidden layers share weight parameters and bias parameters; The first hidden layer includes N hidden units; The nth hidden unit of the mth first hidden layer inputs the nth sequence unit of the feature sequence of the mth node of the graph structure data, and outputs the first update vector, where 1≤m≤M, 1≤n≤N; The mth first extraction layer is used to extract the first update vector output by the Nth hidden unit of the mth first hidden layer as the initial feature of the mth node; That is, the initial features of the node are converted from the feature sequence to the first update vector representation; The second hidden layer inputs graph structure data and outputs an update matrix, where a row vector of the update matrix corresponds to a second update vector of a node; The mth second extraction layer is used to extract the mth row vector of the update matrix and input it into the mth first classifier, and the classification space of the mth first classifier represents the water depth of the grid area that establishes a mapping relationship with the mth node within the second preset time period G; The calculation formula for the nth hidden unit of the mth first hidden layer includes: ; ; ; ; Among them, 1≤u≤12, 1≤v≤12, represents the first update vector output by the nth hidden unit of the mth first hidden layer, The nth sequence unit of the feature sequence of the mth node of the graph structure data input by the nth hidden unit of the mth first hidden layer, express The u-th dimension value of express The upper triangular matrix of express The number of rows and columns of the lower triangular matrix, upper triangular matrix and lower triangular matrix is the same as have the same number of dimensions, The number of dimensions is 12, , , , , and Respectively represent the intermediate vector, the first weight parameter, the second weight parameter, the third weight parameter, the fourth weight parameter and the bias parameter of the nth hidden unit of the mth first hidden layer, sigmoid represents the sigmoid activation function, T represents the transposition operation, represents the first update vector output by the n-1th hidden unit of the mth first hidden layer, Assign a value of 0; The calculation formula for the second hidden layer includes: ; ; ; Where Z represents the update matrix of the second hidden layer output, represents the second update vector of the mth node, represents the set of nodes that build directed edges with the mth node, and represents the first update vector of the mth node and the kth node, and Respectively represent the weight parameters corresponding to the mth node and the kth node, Indicates the weight parameter corresponding to the directed edge from the kth node to the mth node, represents the number of directed edges from the kth node to the mth node, e represents a natural constant, sigmoid represents the sigmoid activation function, Indicates that a stacking operation is performed on the second update vectors of M nodes; The size of the rectangular box, M, R, r and G are all custom parameters, where M represents the number of grid areas, R represents the first preset time period, r represents the preset time interval, and G represents the second preset time period. The size of the rectangular box and M are evaluated and set based on the size of the urban area.
2. The urban flood warning system based on artificial intelligence according to claim 1 is characterized in that: Wind direction is expressed as an angle from 0° to 360°, where 0° or 360° represents north wind, 90° represents east wind, 180° represents south wind, and 270° represents west wind; Slope is expressed as the difference in ground height divided by the horizontal distance over the surface; The landform type is represented by the urbanized area divided by the grid area. The urbanized area includes: road area and building area.
3. The urban flood warning system based on artificial intelligence according to claim 1 is characterized in that: The historical water accumulation depth is used as the sample label of the training sample of the urban flood warning model. Before training the urban flood warning model, the first hidden layer is pre-trained, and the first update vector output by the Nth hidden unit of the first hidden layer is input to the second classifier, and the classification space of the second classifier represents the rainfall at the next time point.
4. The urban flood warning system based on artificial intelligence according to claim 1 is characterized in that: According to the depth of water accumulation in each grid area within the second preset time period G, four warning levels are divided. The first-level warning corresponds to a water accumulation depth between 0 and 10 cm. The corresponding flood prevention and control measures are to issue warning information to remind citizens to pay attention to weather changes. The second-level warning corresponds to a water accumulation depth between 10 cm and 30 cm. The corresponding flood prevention and control measures are to strengthen monitoring and patrols, and report water accumulation in a timely manner. The third-level warning corresponds to a water accumulation depth between 30 cm and 50 cm. The corresponding flood prevention and control measures are to remind citizens to pay attention to safety and avoid going to waterlogged areas. The fourth-level warning corresponds to a water accumulation depth of more than 50 cm. The corresponding flood prevention and control measures are to issue emergency notices to remind citizens to stay away from waterlogged areas.
5. An urban flood warning method based on artificial intelligence, characterized in that: Executing an artificial intelligence-based urban flood warning system as described in any one of claims 1 to 4 comprises the following steps: Step S201, dividing the urban area into M grid areas by rectangular frames; Step S202, within a first preset time period R, collecting meteorological data, hydrological data and geographical data of M grid areas at preset time intervals r, and generating a feature sequence for each grid area; Step S203, constructing graph structure data according to the feature sequences of the M grid areas; Step S204, inputting the graph structure data into the urban flood warning model, and outputting the water depth of the M grid areas within the second preset time period G; Step S205: generating flood prevention and control measures according to the water depths of the M grid areas within the second preset time period G.
