Intelligent treatment methods, systems and equipment for urban waterlogging

Abstract

The present invention provides a method, system and equipment for intelligent handling of urban waterlogging, which include: constructing a feature vector and a knowledge graph of a current waterlogging event based on multimodal monitoring data; selecting similar historical cases based on feature vector similarity and knowledge graph structure similarity; scoring the handling effects achieved by similar historical cases after adopting corresponding response strategies, and obtaining a set of recommended response strategies; constructing an urban waterlogging simulation model, assigning attributes to each grid node in the model and configuring the position of each drainage device in the model; generating a dynamic deployment logic for drainage equipment based on the selected optimal response strategy and the position of each drainage device in the model, obtaining an urban waterlogging simulation deduction model and deducing simulation prediction results based on the model; performing prevention and control measures based on the dynamic deployment logic of drainage equipment, and obtaining actual waterlogging status monitoring results; and determining the effectiveness of the current optimal response strategy based on the simulation prediction results and the actual waterlogging status monitoring results.

Description

Technical Field

[0001] The present invention relates to the field of urban intelligent operation and maintenance technology, and in particular to a method, system and equipment for intelligent disposal of urban waterlogging. Background Art

[0002] With the acceleration of urbanization and the frequent occurrence of extreme weather events, urban flood control and management face severe challenges. Traditional urban flood management systems rely on single-sensor data or manual experience to achieve flood control. These systems suffer from inefficient multimodal data fusion, insufficient use of historical case studies, a disconnect between simulation and real-time feedback, and delayed knowledge updates. These issues lead to insufficient prediction accuracy, poor strategy adaptability, and delayed response. While existing technologies incorporate the Internet of Things and artificial intelligence, data fragmentation, static knowledge graphs, and fixed simulation parameters still hinder global perception and dynamic optimization capabilities.

[0003] In response to key technical issues in current urban waterlogging management systems, such as insufficient data integration, weak knowledge systems, limited simulation capabilities, and a lack of feedback optimization mechanisms for policy execution, there is an urgent need for an intelligent, closed-loop waterlogging prevention and control system that can integrate multi-source data, correlate historical experience, calibrate strategies in real time, and continuously evolve the knowledge base. Summary of the Invention

[0004] In view of the above problems, the present invention is proposed to provide a method, system and device for intelligent disposal of urban waterlogging that overcome the above problems or at least partially solve the above problems.

[0005] One aspect of the present invention provides an intelligent method for handling urban waterlogging, the method comprising:

[0006] Construct the feature vector and knowledge graph of the current waterlogging event based on the multimodal monitoring data of the current waterlogging event;

[0007] Based on the feature vector similarity and knowledge graph structure similarity, historical cases similar to the current flooding incident are selected from the preset historical case data set to obtain a similar case set;

[0008] Based on the preset treatment effect evaluation indicators, the treatment effects achieved by each historical case in the similar case set after adopting the corresponding response strategy are scored, and the recommended set of response strategies that match the current waterlogging incident is selected based on the scores;

[0009] Constructing an urban waterlogging simulation model, wherein the urban waterlogging simulation model divides the current urban area into grids, assigns attribute information to each grid node, and configures the position of each drainage device in the urban waterlogging simulation model according to pipe network distribution data in a preset urban model;

[0010] The highest-scoring response strategy in the recommended response strategy set is selected as the optimal response strategy. Based on the optimal response strategy and the location of each drainage device in the urban waterlogging simulation model, a dynamic deployment logic for drainage equipment is generated to obtain an urban waterlogging simulation deduction model. The predicted water depth at each grid node when a specified end time is reached is predicted based on the urban waterlogging simulation deduction model to obtain a simulation prediction result corresponding to the optimal response strategy.

[0011] Conduct prevention and control measures in the current urban area based on the dynamic deployment logic of drainage equipment to obtain actual waterlogging monitoring results;

[0012] The effectiveness of the current optimal response strategy is determined based on the simulation prediction results and the actual waterlogging monitoring results. If it is effective, the current optimal response strategy will continue to be implemented. Otherwise, the optimal response strategy will be reselected to deal with the current waterlogging event.

[0013] Another aspect of the present invention provides an intelligent urban waterlogging treatment system, comprising:

[0014] Multimodal data fusion module, used to construct the feature vector and knowledge graph of the current waterlogging event based on the multimodal monitoring data of the current waterlogging event;

[0015] The graph reasoning module is used to select historical cases similar to the current flooding incident from a preset historical case dataset based on feature vector similarity and knowledge graph structure similarity to obtain a set of similar cases;

[0016] The strategy recommendation module is used to score the treatment effects achieved by each historical case in the similar case set after adopting the corresponding response strategy based on the preset treatment effect evaluation index, and select the recommended set of response strategies that match the current waterlogging incident based on the score;

[0017] A simulation model construction module is used to construct an urban waterlogging simulation model, which divides the current urban area into grids, assigns attribute information to each grid node, and configures the location of each drainage device in the urban waterlogging simulation model based on the pipe network distribution data in the preset urban model;

[0018] A prediction and deduction module is used to select the response strategy with the highest score in the recommended response strategy set as the optimal response strategy, generate dynamic deployment logic for drainage equipment based on the optimal response strategy and the location of each drainage equipment in the urban waterlogging simulation model, obtain the urban waterlogging simulation and deduction model, and predict the water depth of each grid node at the specified end time based on the urban waterlogging simulation and deduction model to obtain the simulation prediction result corresponding to the optimal response strategy;

[0019] The actual measurement acquisition module is used to carry out prevention and control measures in the current urban area based on the dynamic deployment logic of drainage equipment and obtain the actual waterlogging monitoring results;

[0020] The strategy execution detection module is used to determine whether the current optimal response strategy is effective based on the simulation prediction results and the actual waterlogging monitoring results. If it is effective, the current optimal response strategy will continue to be executed; otherwise, the optimal response strategy will be reselected to deal with the current waterlogging event.

[0021] Another aspect of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor; when the computer program is executed by the processor, the steps of the above-mentioned intelligent urban waterlogging disposal method are implemented.

[0022] Another aspect of the present invention provides a computer program product having a computer program stored thereon, which implements the steps of the above-mentioned intelligent urban waterlogging treatment method when executed by a processor.

[0023] The embodiments of the present invention provide an intelligent method, system, and device for handling urban waterlogging, and propose an intelligent decision-making and handling system that integrates multimodal perception, knowledge graph reasoning, dynamic simulation verification, and closed-loop self-correction capabilities. The system aims to achieve real-time fusion of multi-source data, high-precision scene restoration, intelligent policy reasoning and recommendation, dynamic optimization of policy execution, and continuous evolution of the knowledge graph, build a closed-loop management system, comprehensively improve the emergency response efficiency and governance capabilities of cities under extreme climates for waterlogging, solve the problems of data isolation, policy rigidity, and delayed response, and meet the technical and market needs for efficient governance of urban waterlogging.

[0024] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are specifically listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present invention. The same reference symbols are used throughout the drawings to represent the same components. In the drawings:

[0026] Figure 1 A flow chart of an intelligent urban waterlogging treatment method according to an embodiment of the present invention;

[0027] Figure 2A structural block diagram of the urban waterlogging intelligent treatment system proposed in an embodiment of the present invention. DETAILED DESCRIPTION

[0028] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0029] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a," "an," "said," and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the stated features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0030] Those skilled in the art will understand that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by those skilled in the art in the art to which this invention belongs. It should also be understood that terms such as those defined in common dictionaries should be understood to have meanings consistent with their meanings in the context of the prior art and, unless specifically defined, should not be interpreted in an idealized or overly formal sense.

[0031] Figure 1 The flowchart of the intelligent treatment method for urban waterlogging according to one embodiment of the present invention is schematically shown. Figure 1 The urban waterlogging intelligent treatment method according to the embodiment of the present invention specifically includes the following steps:

[0032] S11. Construct the feature vector and knowledge graph of the current waterlogging event based on the multimodal monitoring data of the current waterlogging event.

[0033] Specifically, the present invention uses rain gauges, lidar, pipe pressure sensors, and water depth sensors to acquire real-time multimodal monitoring data corresponding to current urban flooding events, including precipitation intensity, topography, pipe network flow, and water depth data. Using an attention mechanism, the obtained data from different modalities is mapped into a unified semantic space to generate a fused feature vector for the current flooding event. Furthermore, the present invention extracts entities and relationships from text files related to the current flooding event to form structured entity-relationship-entity triples. This structured triple data is then transformed into a knowledge graph to generate a knowledge graph for the current flooding event. This knowledge graph contains entities related to the current flooding event (e.g., disaster events, situational features) and their relationships. The text file for the current flooding event consists of data on precipitation intensity, topography, pipe network flow, and water depth, along with spatiotemporal coordinates and a semantic description of the event.

[0034] S12. Select historical cases similar to the current waterlogging incident from a preset historical case dataset based on feature vector similarity and knowledge graph structure similarity to obtain a similar case set.

[0035] In this embodiment, historical flood response case text data is collected from urban management departments, meteorological departments, news media, and other channels. This data includes information such as the time, location, cause, scope of impact, response measures taken, and final results of flooding. This data is then used to construct a historical case dataset. Leveraging these historical flood response cases, a structured knowledge graph and feature vectors for each historical case are constructed, enabling case-based reasoning and strategy recommendations based on the current situation.

[0036] S13. Score the treatment effect achieved by each historical case in the similar case set after adopting the corresponding response strategy based on the preset treatment effect evaluation index, and select a recommended set of response strategies that matches the current waterlogging incident based on the score.

[0037] S14. Construct an urban waterlogging simulation model, which divides the current urban area into grids, assigns attribute information to each grid node, and configures the position of each drainage device in the urban waterlogging simulation model according to the pipe network distribution data in the preset urban model.

[0038] The attribute information of each grid node includes the center coordinates, area, precipitation intensity, elevation value, water depth, whether there is drainage equipment, and the location and flow of the drainage equipment.

[0039] The urban model M includes data such as land use type data, topographic data (used to generate DEM), building and road information (including building boundaries, road networks, their types and widths), pipe network distribution data (location and flow of drainage equipment, etc.), and urban physical geometry information (used to calculate grid area and interface length).

[0040] S15. Select the response strategy with the highest score in the response strategy recommendation set as the optimal response strategy, generate the dynamic deployment logic of the drainage equipment according to the optimal response strategy and the position of each drainage equipment in the urban waterlogging simulation model, obtain the urban waterlogging simulation deduction model, and predict the water depth prediction value of each grid node when the specified end time is reached according to the urban waterlogging simulation deduction model to obtain the simulation prediction result corresponding to the optimal response strategy.

[0041] In this embodiment, the pipe network distribution data of the city model M provides the physical existence of the drainage equipment. In step S14, the location of each drainage device in the urban waterlogging simulation model has been determined based on the pipe network distribution data of the city model M. The optimal response strategy P1 is used to provide dynamic deployment logic for drainage equipment, such as which equipment to enable and the location of new equipment. Based on P1, the location of drainage equipment is filtered or adjusted from the pipe network data of M, such as adding a new pump station, shutting down faulty equipment, etc., to obtain an urban waterlogging simulation deduction model. For example, the optimal response strategy P1 provides for adding drainage at key intersections, and then extracts a list of coordinates of all drainage wells or pump stations from the city model M. The recommended deployment area is spatially matched with the pipe network nodes to determine the coordinates of the enabled drainage equipment.

[0042] S16. Conduct prevention and control measures in the current urban area based on the dynamic deployment logic of drainage equipment to obtain actual waterlogging monitoring results.

[0043] S17. Determine whether the current optimal response strategy is effective based on the simulation prediction results and the actual waterlogging monitoring results. If effective, continue to execute the current optimal response strategy. Otherwise, reselect the optimal response strategy to handle the current waterlogging event.

[0044] The intelligent urban waterlogging handling method provided by the embodiment of the present invention constructs a feature vector and a knowledge graph of the current waterlogging event based on multimodal monitoring data; selects similar historical cases based on the similarity of the feature vectors and the similarity of the knowledge graph structure; scores the handling effects achieved by similar historical cases after taking corresponding response strategies, and obtains a recommended set of response strategies: constructs an urban waterlogging simulation model, assigns attributes to each grid node in the model and configures the position of each drainage device in the model; generates a dynamic deployment logic for drainage equipment based on the selected optimal response strategy and the position of each drainage device in the model, obtains an urban waterlogging simulation deduction model, and deduces simulation prediction results based on this; performs prevention and control according to the dynamic deployment logic of drainage equipment, and obtains actual waterlogging status monitoring results; and determines the effectiveness of the current optimal response strategy based on the simulation prediction results and the actual waterlogging status monitoring results.

[0045] This invention proposes an intelligent decision-making and disposal system that integrates multimodal perception, knowledge graph reasoning, dynamic simulation verification and closed-loop self-correction capabilities. It aims to achieve real-time fusion of multi-source data, high-precision scene restoration, intelligent policy reasoning and recommendation, dynamic optimization of policy execution and continuous evolution of knowledge graphs, build a closed-loop management system, comprehensively improve the city's emergency response efficiency and governance capabilities under extreme climates, solve the problems of data isolation, policy rigidity and response lag, and meet the technical and market needs for efficient urban waterlogging governance.

[0046] In the embodiment of the present invention, step S11 constructs the feature vector of the current waterlogging event based on the multimodal monitoring data of the current waterlogging event, and specifically includes the following steps (not shown in the figure):

[0047] S111. Obtain the precipitation intensity, topography, pipe network flow, and water depth of the current flooding event, and perform normalization and time alignment preprocessing on the acquired data. Specific normalization and time alignment preprocessing methods can be implemented using existing technologies and are not specifically limited in the present invention.

[0048] S112. Extract feature data from the preprocessed precipitation intensity, terrain, pipe network flow, and waterlogging depth to obtain a precipitation intensity feature vector, a terrain feature vector, a pipe network flow feature vector, and a waterlogging depth feature vector. Specifically, for precipitation intensity data, Fourier transform is used to extract its frequency domain features to obtain a precipitation intensity feature vector. For terrain data, three-dimensional point cloud processing technology (PCL library) is used to extract features such as terrain undulation and slope to obtain a terrain feature vector. For pipe network flow data, wavelet analysis is used to extract the time-frequency characteristics of flow changes to obtain a pipe network flow feature vector. For waterlogging depth data, sliding window averaging is used to extract its trend features to obtain a waterlogging depth feature vector.

[0049] S113, based on the attention mechanism, feature fusion is performed on the precipitation intensity feature vector, terrain feature vector, pipe network flow feature vector, and water depth feature vector, and the obtained fusion feature vector is used as the feature vector of the current waterlogging event. Further, step S113 specifically includes: splicing the precipitation intensity feature vector, terrain feature vector, pipe network flow feature vector, and water depth feature vector into a long vector in the column direction; constructing a learnable weight matrix Based on the weight matrix, a linear transformation is performed on the spliced ​​long vector to obtain an attention score vector Z, and the attention score vector is converted into an attention weight vector through the Softmax function to obtain the attention weights of the precipitation intensity feature vector, the terrain feature vector, the pipe network flow feature vector and the water depth feature vector when performing feature fusion; the precipitation intensity feature vector, the terrain feature vector, the pipe network flow feature vector and the water depth feature vector are weightedly spliced ​​according to their corresponding attention weights to obtain a fused feature vector.

[0050] The present invention introduces the attention mechanism to dynamically weight the above-mentioned real-time precipitation intensity, terrain, pipe network flow, and water depth data feature vectors. First, the four feature vectors are spliced ​​into a long vector in the column direction, and then a learnable weight matrix is ​​used. Perform a linear transformation to obtain the attention score vector Z, and finally apply the Softmax function to obtain the attention weights of the four modalities. The formula is as follows:

[0051] ,

[0052] Where F is a 4D column vector; the semicolon indicates vertical splicing; is the terrain feature vector; is the precipitation intensity eigenvector; is the pipe network flow characteristic vector; is the water depth feature vector. d refers to the length of a single feature vector, such as the terrain feature 20 parameters are extracted, so d is equal to 20. In this embodiment, the dimensions of the four types of features must be consistent, all being d; T is the transpose of the matrix. Indicates that the vector is 4d rows and one column;

[0053] ,

[0054] in, is a learnable reference matrix, and the obtained , Represents a matrix with 4 rows and 4d columns. is a 4-row, 4d-column matrix. The 4 rows correspond to the 4 eigenmodes: terrain, precipitation, pipe network, and waterlogging, and the 4d-column is the dimension of the concatenated eigenvector. It is obtained through model training. First, All elements of are initialized to random values ​​(for example, d = 50, then It is a 4*200 matrix), and then goes through forward propagation, loss calculation, back propagation and parameter update, and iteration until convergence.

[0055] ,

[0056] get:

[0057] ;

[0058] ;

[0059] ;

[0060] ,

[0061] in, 、 、 、 are the attention weights of the four modalities, representing the importance of the corresponding feature vectors in the fusion process, and Z is a 4-dimensional column vector, and Z1, Z2, Z3, and Z4 are the first, second, third, and fourth rows of the column vector respectively.

[0062] Finally, the multimodal feature vectors are multiplied by the corresponding weights and concatenated according to the attention weights to obtain a fused feature vector of unified dimension.

[0063] ,

[0064] in, is the output fusion feature vector; Represents the concatenation operation of feature vectors, which is used to obtain a fusion vector of uniform dimension.

[0065] In this embodiment of the present invention, step S12 selects historical cases similar to the current flooding event from a preset historical case dataset based on feature vector similarity and knowledge graph structure similarity, and specifically includes the following steps (not shown in the figure):

[0066] S121. Label the entities and relationships of the text data in the historical case data set, where the entities include disaster event entities, situation feature entities, disposal measure entities, and response result entities, and the relationships represent the association relationships between different entities; extract the entities and relationships in the text data to form structured triples of entity-relationship-entity, and semantically store the structured triple data through a knowledge graph to obtain a historical case knowledge graph.

[0067] Specifically, this method first preprocesses the text data in the historical case dataset through data cleaning, word segmentation, and part-of-speech tagging. Entity extraction, relationship extraction, and knowledge graph storage are then performed sequentially to construct a historical case knowledge graph for urban flooding management strategies and effectiveness. Entity types include: disaster events (e.g., "sudden rainstorm," "flooding warning"), situational characteristics (e.g., "low-lying terrain," "inadequate drainage capacity"), management measures (e.g., "dispatching pumping and drainage," "traffic closure," "evacuation guidance"), and response results (e.g., "flooding relief," "traffic restoration"). Relation types include "cause is," "needs to respond to," "effect is," and "applicable to." Attribute information includes measures such as "response time," "duration," "applicable areas," "trigger conditions," "cost-benefit assessment," and "feedback level." Finally, the graph is organized as structured triples (entity-relationship-entity) and stored in a graph database (e.g., Neo4j), forming a multi-layered urban flooding knowledge network centered around the "disaster-response-effectiveness" framework.

[0068] S122. Construct a corresponding feature vector for each historical case in the historical case dataset.

[0069] Specifically, the TransE algorithm can be used to construct a feature vector of the corresponding historical case based on the entities and relationships extracted from each historical case.

[0070] S123. Calculate the similarity between the feature vectors of each historical case and the feature vector of the current waterlogging event.

[0071] S124. Extract the sub-graph of each historical case from the historical case knowledge graph, and calculate the structural similarity between the sub-graph of each historical case and the knowledge graph of the current flooding incident.

[0072] In the embodiment of the present invention, the graph edit distance between the sub-knowledge graph of each historical case and the knowledge graph of the current flooding event can be calculated based on the A* algorithm, and the structural similarity of the knowledge graphs can be determined based on the graph edit distance. The graph edit distance calculation formula is as follows:

[0073] ,

[0074] in, is the graph edit distance (calculated based on the A* algorithm); The knowledge graph of the current flooding event includes the entities of the current flooding event (such as disaster events, situation characteristics) and their relationships; is the subgraph of the i-th historical case, which contains the entities and their relationships of the i-th historical case; for The node that needs to be deleted; To delete a node the costs required; for The node to be inserted (i.e. missing nodes in ); To insert a node the costs required; For the general Nodes in Replace with Nodes in the costs required; is the edge that needs to be edited in the graph; For the opposite side The cost of performing the operation.

[0075] S125. Calculate the event similarity between each historical case and the current flooding event based on the feature vector similarity, knowledge graph structure similarity, and time decay factor. The event similarity calculation formula is as follows:

[0076] ,

[0077] in, is the event similarity, W, V, and U are the weight coefficients corresponding to the feature vector similarity, knowledge graph structure similarity, and time decay factor, respectively, satisfying , the default settings are W=0.6; V=0.2, U=0.2, which can be adjusted according to actual application; is the characteristic vector of the current waterlogging event; is the feature vector of the historical case; is the L2 norm of the vector (i.e., the length of the vector); The knowledge graph of the current flooding event includes the entities and relationships of the current flooding event; is the subgraph of the i-th historical case, including the entities and relationships of the historical case, for and The similarity of knowledge graph structure; The default value of the time decay coefficient is 0.02 (unit: per hour), which is used to control the speed of time decay; The timestamp of the current flooding event; The timestamp of the historical case.

[0078] S126. Select historical cases whose event similarity is greater than a preset similarity threshold from a preset historical case data set to obtain historical cases similar to the current waterlogging event. The specific formula is as follows:

[0079] ,

[0080] in, A collection of similar cases. It is a specific historical case that is highly similar to the current waterlogging situation; is the similarity threshold, optional, and is set to 0.75 by default.

[0081] In the embodiment of the present invention, step S13 scores the treatment effect achieved by each historical case in the similar case set after adopting the corresponding response strategy according to the preset treatment effect evaluation index, which specifically includes:

[0082] The weights of various treatment effect evaluation indicators are configured. In a specific example, the treatment effect evaluation indicators include but are not limited to considering one or more of response speed, implementation cost, and disaster reduction effect.

[0083] The score of the disposal effect achieved by each historical case after taking the corresponding response strategy is calculated based on the preset scoring function.

[0084] In this example, we evaluate and rank the responses to historical cases extracted from a collection of similar cases based on the actual results achieved after implementing the corresponding measures. Specifically, we use a weighted scoring method, assigning weights to different responses based on the effectiveness of the measures. Ultimately, the responses with the highest scores are selected as the recommended response strategies for the current flooding situation. The scoring function is as follows:

[0085] ;

[0086] ,

[0087] in, is the weight of the j-th treatment effect evaluation indicator, reflecting its importance; This is the preset magnification factor. The default value is 1 and can be adjusted according to the actual scenario and needs. The larger the value, the greater the weight of the more important indicators will be, and the smaller the value, the more evenly the weight distribution will be. is the importance score of the jth treatment effect evaluation index, which can be calculated using the AHP (Analytical Hierarchy Process) method; m is the total number of treatment effect evaluation indicators; g is the temporary index of the summation operation; is the importance score of the g-th treatment effect evaluation indicator; For strategy The final weighted total score; For strategy The score on the j-th treatment effect evaluation indicator; is the i-th candidate response strategy.

[0088] Through the scoring function After evaluation and sorting, the optimal response strategy set for the current flooding incident is obtained: .

[0089] The embodiment of the present invention can be based on the real-time precipitation intensity obtained ,terrain , pipe network flow , water depth data , combining the city's corresponding land use type, terrain and landform data, and building and road information into an urban model Assign attributes to the urban waterlogging simulation model.

[0090] Specifically, step S14 assigns a value to the attribute information of each grid node and configures the position of each drainage device in the urban waterlogging simulation model according to the pipe network distribution data in the preset urban model, including:

[0091] (1) According to the precipitation intensity, terrain, and water depth data, each grid node in the urban waterlogging simulation model is configured with rainfall intensity, elevation value, and water depth. , which can be processed into a gridded rainfall intensity field, and each grid point (x, y) divided into urban areas is assigned a rainfall intensity R(x, y), accurately presenting the spatial distribution of rainfall; topography In terms of the above, the topographic data in the city model M is combined to generate a digital elevation model (DEM), and the grid point (x, y) is assigned an elevation value Z(x, y). At the same time, the land use type and building road information in M ​​are used to modify the DEM, such as marking the building area as a non-floodable area and giving special treatment to the road area; the water depth As part of the initial water depth field, it is used for initialization and the water depth values ​​of the monitoring points are interpolated onto the grid in combination with the terrain data in the urban model M. In the urban model M, the land use type is used to determine the diffusion rate d in the future. The fusion forms a more refined DEM, and the building and road information is used to generate the building boundary mask (presented in the form of a binary image, with buildings as 1 and non-buildings as 0) and the road network (marking the road type and width). Specifically including: using The initial water depth of each grid is calculated based on the monitoring point data and the elevation information provided by the digital elevation model. For the building area, due to its non-floodable characteristics, its water depth is forced to be 0, that is, when the building boundary mask hour, For non-building areas, the initial water depth is the interpolation result, that is, when hour, .

[0092] (2) According to the physical geometry information in the preset city model, the center coordinates and area of ​​each grid node in the urban flooding simulation model are configured. Specifically, the urban flooding simulation model divides the current urban area into grids (x, y), introduces building, road and land use boundary conditions, and assigns the center coordinates of each grid unit i according to the physical geometry information in the city model. and area The grid size is not fixed, but is flexibly adjusted according to the building and road information: in densely built areas, in order to more accurately simulate water flow changes, the grid is encrypted, for example, set to 1m×1m; in open areas, the grid can be appropriately sparse, such as 5m×5m.

[0093] The boundary conditions are specifically set as follows: in terms of building boundaries, in the subsequent water flow calculation process, the building area grid does not participate in the calculation (that is, the water depth value is not updated), and the water flow of adjacent grids cannot pass through the building boundary; in terms of road boundaries, considering the special impact of roads on water flow diffusion, the water flow diffusion rate in the road area will be multiplied by an enhancement factor, which can be set to 1.6. This adjustment will be reflected in the subsequent diffusion rate d calculation. Finally, after the initialization of the grid, the properties of each grid include the initial water depth ,area and elevation .

[0094] (3) According to the pipe network flow and the pipe network distribution data in the city model, the location of each drainage device in the urban waterlogging simulation model and the flow and influence radius of the corresponding drainage device are configured. Specifically, for the pipe network flow , combined with the pipe network distribution data in the city model M to determine the location coordinates of each drainage device and the corresponding flow (From ), and estimate the influence radius of each drainage device based on the relevant pipe network distribution data such as pipe network density and equipment power in the urban model M For example, the greater the pipe network density, the wider the impact range. The greater the power of the equipment, the greater the pumping and drainage capacity, and the wider the impact range.

[0095] In the embodiment of the present invention, the final urban waterlogging simulation model can be obtained based on the two-dimensional hydrodynamic model LISFLOOD-FP, and the formula is as follows:

[0096] ,

[0097] Furthermore, the step S15 of predicting the water depth of each grid node when the designated end time is reached according to the urban waterlogging simulation model specifically includes:

[0098] The urban flooding simulation model is used to predict the evolution of waterlogging to construct a continuous shallow water dynamics differential equation. Specifically, the core control equation of the LISFLOOD-FP model (the simplified shallow water equation) is used in combination with strategic intervention to predict the evolution of waterlogging. The continuous shallow water dynamics differential equation used to describe the dynamic change of water depth over time and space is obtained as follows:

[0099] ;

[0100] ;

[0101] ,

[0102] in, is the change in water depth per unit time for each grid node; is the divergence operator; d is the water diffusion rate; is the spatial gradient of water accumulation; h is the depth of water accumulation; is the current optimal response strategy at grid node i The net rate of change caused by The number of drainage equipment; is the flow rate of the kth drainage device; is the influence radius of the kth drainage device; is the center coordinate of grid node i; is the deployment location coordinate of the kth drainage device; For net rain strength; is the measured rainfall intensity at position (x, y); is the loss coefficient related to land use type L, which indicates the proportion of rainfall lost due to surface interception, infiltration, etc.;

[0103] Among them, the water diffusivity d is related to the land use type L and the current water depth h, and the calculation formula is: ,in, is the land type coefficient, which represents the coefficient of surface water transmission capacity and can be dynamically obtained from hydrological manuals, local drainage design specifications or field infiltration experiments. α is the water depth index, which is usually fixed at 2 / 3 (the theoretical value of the Manning formula in fluid mechanics).

[0104] The explicit finite difference method is used to discretize the continuous shallow water dynamics differential equations and convert them into a computable discrete form for time step calculation. Iteratively update the water depth value of each grid node to obtain the discretized water depth prediction formula as follows:

[0105] ,

[0106] in, ,

[0107] in, is the water depth of grid node i at time step n+1; is the water depth of grid node i at time step n; is the time step (unit: s); is the area of ​​grid node i; is the set of adjacent grid nodes of grid node i; is the diffusion rate at the interface between grid node i and adjacent grid node j; is the interface length between grids i and j, i.e., the actual physical length of the shared boundary between adjacent grid cells; is the grid cell center distance; At the grid node i The net rate of change caused by the current optimal response strategy;

[0108] When the urban waterlogging simulation model reaches the specified end time, the predicted water depth value h(x, y, T) of each grid node is obtained according to the water depth prediction formula, which is the urban waterlogging map obtained by simulation prediction. .

[0109] .

[0110] In the embodiment of the present invention, judging whether the current optimal response strategy is effective based on the simulation prediction results and the actual waterlogging monitoring results in step S17 includes:

[0111] Calculate the absolute water depth error and relative error rate between the predicted water depth value of each grid node in the simulation prediction results when executing the current optimal response strategy and the water depth monitoring value of the actual monitoring point corresponding to each grid node. The formula is as follows:

[0112] ;

[0113] ,

[0114] in, is the absolute water depth error at the current moment; N is the number of monitoring points; is the predicted value of water depth; is the monitoring value of water depth; is the relative error rate; To prevent division by zero constant;

[0115] The effectiveness of the current optimal response strategy is determined based on the absolute water depth error and relative error rate. If both the absolute water depth error and the relative error rate are less than the preset error tolerance threshold, the current optimal response strategy is determined to be effective. The specific judgment principles are as follows:

[0116] ,

[0117] in, As a result of judgment. The currently recommended strategy set; is the error tolerance threshold.

[0118] if and , it means that the simulation prediction result is consistent with the actual flooding situation, and the current optimal response strategy is determined Valid, you can continue to execute the strategy Otherwise, it means that the simulation prediction results are inconsistent with the actual waterlogging situation. In this case, the strategy correction is triggered, and the optimal response strategy is re-selected through dynamic self-correction iteration to deal with the current waterlogging incident.

[0119] Furthermore, the specific implementation process of reselecting the optimal response strategy through dynamic self-correction iteration is as follows:

[0120] From the obtained set of recommended coping strategies Excluding the current optimal response strategy , for the remaining candidate recommendation strategy set The strategies in the simulation are simulated again in sequence, and the absolute water depth error of each strategy is calculated first. and relative error rate , according to the absolute depth error of each response strategy and relative error rate Calculate the comprehensive error of the response strategy using the following formula:

[0121] ,

[0122] in, For coping strategies The comprehensive error of is the weight coefficient of absolute water depth error; are the weight coefficients of the relative error rate, and the default weights are 0.7 and 0.3 respectively.

[0123] From the remaining candidate strategy set In the equation, the response strategy with the smallest comprehensive error is selected as the new optimal response strategy. The formula is as follows:

[0124] ,

[0125] in, For the new optimal response strategy; To take the strategy that minimizes the comprehensive error.

[0126] In one embodiment of the present invention, the intelligent urban flooding management method provided by the present invention also includes a knowledge base incremental update step. Using this incremental update method, newly added urban flooding cases and real-time policy execution records are added to the existing knowledge graph. Specifically, the following steps (not shown in the accompanying figure) are included:

[0127] S18. Taking the current flooding incident and the corresponding optimal response strategy as a new case, performing entity extraction and relationship extraction on the new case, and forming candidate structured triples of entity-relationship-entity;

[0128] S19. When the similarity between the candidate entity extracted from the case to be added and all the existing entities in the historical case knowledge graph is less than the preset entity similarity judgment threshold, the candidate entity is added to the existing entity set to realize the incremental update of the knowledge base; when the similarity between the candidate entity extracted from the case to be added and all the existing entities in the historical case knowledge graph is less than the preset entity similarity judgment threshold, and when the candidate structured triple is different from the existing triple relationship set in the historical case knowledge graph, the candidate structured triple is added to the existing relationship set to realize the incremental update of the knowledge base.

[0129] In this embodiment, entity extraction and relationship extraction are performed on new cases. When the similarity between the candidate entity and all existing entities is less than an entity similarity judgment threshold, it will be added to the existing entity set. When the triple does not exist in the existing relationship set at all, the new relationship will be added to the relationship set. The specific formula is as follows:

[0130] ;

[0131] ,

[0132] in, A collection of entities to be added; The candidate entities are extracted from the newly added case text; is the entity extraction result of the new text; e is the entity in the existing knowledge base; is the similarity between the candidate entity and the existing entity, which can be calculated using the cosine similarity calculation formula; is the entity similarity determination threshold, which can be selected as 0.9 by default; E is the existing entity set in the knowledge base; is the relationship triple to be added; is a candidate structured triple; is the relation extraction result of the new case text; R is the existing relation set in the knowledge base.

[0133] Finally, we get the incrementally updated knowledge graph ,in, Add updated attribute fields to each policy node.

[0134] This invention boasts the advantages of an intelligent closed-loop system that integrates multimodal perception, knowledge graph reasoning, simulation feedback verification, and incremental learning. It dynamically integrates multi-source data through an attention mechanism, significantly improving the accuracy and completeness of urban flood situation awareness. It leverages structured knowledge graphs and similarity reasoning to implement intelligent strategy recommendations for different scenarios. It introduces an error calibration mechanism between simulation prediction and real-time observation to support dynamic self-correction of strategies. It also incorporates the FUP algorithm to continuously update and evolve the knowledge base, overcoming the shortcomings of existing systems such as data fragmentation, static knowledge, and rigid strategies, effectively enhancing the intelligence, reliability, and adaptability of urban flood emergency response.

[0135] For simplicity of description, the method embodiments are described as a series of actions. However, those skilled in the art should be aware that the embodiments of the present invention are not limited by the order of the actions described, because certain steps can be performed in other orders or simultaneously according to the embodiments of the present invention. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present invention.

[0136] Another embodiment of the present invention further provides an intelligent urban waterlogging treatment system, which includes a functional module for implementing any of the above-mentioned intelligent urban waterlogging treatment methods. Figure 2 The following schematically shows a structural block diagram of an intelligent urban waterlogging treatment system according to another embodiment of the present invention. Figure 2 The urban waterlogging intelligent treatment system of this embodiment specifically includes a multimodal data fusion module 201, a graph reasoning module 202, a strategy recommendation module 203, a simulation model construction module 204, a prediction and deduction module 205, a measurement acquisition module 206, and a strategy execution detection module 207, wherein:

[0137] The multimodal data fusion module 201 is used to construct a feature vector and a knowledge graph of the current waterlogging event based on the multimodal monitoring data of the current waterlogging event;

[0138] A graph reasoning module 202 is configured to select historical cases similar to the current flooding incident from a preset historical case dataset based on feature vector similarity and knowledge graph structure similarity to obtain a similar case set;

[0139] The strategy recommendation module 203 is used to score the treatment effect achieved by each historical case in the similar case set after adopting the corresponding response strategy based on the preset treatment effect evaluation index, and select the recommended response strategy set that matches the current waterlogging incident based on the score:

[0140] A simulation model construction module 204 is configured to construct an urban waterlogging simulation model, wherein the urban waterlogging simulation model divides the current urban area into grids, assigns attribute information to each grid node, and configures the location of each drainage device in the urban waterlogging simulation model based on pipe network distribution data in a preset urban model;

[0141] The prediction and deduction module 205 is configured to select the highest-scoring response strategy from the recommended response strategy set as the optimal response strategy, generate dynamic drainage device deployment logic based on the optimal response strategy and the location of each drainage device in the urban waterlogging simulation model, obtain the urban waterlogging simulation and deduction model, and predict the water depth at each grid node at a specified end time based on the urban waterlogging simulation and deduction model to obtain a simulation prediction result corresponding to the optimal response strategy;

[0142] The actual measurement acquisition module 206 is used to carry out prevention and control measures for the current urban area according to the dynamic deployment logic of drainage equipment and obtain the actual waterlogging monitoring results;

[0143] The strategy execution detection module 207 is used to determine whether the current optimal response strategy is effective based on the simulation prediction results and the actual waterlogging monitoring results. If it is effective, the current optimal response strategy will continue to be executed; otherwise, the optimal response strategy will be re-selected to deal with the current waterlogging event.

[0144] The urban waterlogging intelligent treatment system provided by the embodiment of the present invention further includes a dynamic self-modification module, which is used to: Excluding the current optimal response strategy , for the remaining candidate recommendation strategy set The strategies in the simulation are simulated again in sequence, and the absolute water depth error of each strategy is calculated first. and relative error rate , according to the absolute depth error of each response strategy and relative error rate Calculate the comprehensive error of the response strategy; from the remaining candidate strategy set Among them, the response strategy with the smallest comprehensive error is selected as the new optimal response strategy.

[0145] The intelligent urban waterlogging disposal system provided by an embodiment of the present invention also includes a knowledge base incremental update module. The knowledge base incremental update is used to take the current waterlogging event and the corresponding optimal response strategy as new cases to be added, perform entity extraction and relationship extraction on the new cases to be added, and form candidate structured triples of entity-relationship-entity; when the similarity between the candidate entity extracted from the new case and all existing entities in the historical case knowledge graph is less than a preset entity similarity judgment threshold, the candidate entity is added to the existing entity set to achieve incremental update of the knowledge base; when the similarity between the candidate entity extracted from the new case and all existing entities in the historical case knowledge graph is less than a preset entity similarity judgment threshold, and when the candidate structured triple is different from the existing triple relationship set in the historical case knowledge graph, the candidate structured triple is added to the existing relationship set to achieve incremental update of the knowledge base.

[0146] As for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0147] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0148] In addition, another embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor; when the computer program is executed by the processor, the steps of the above-mentioned intelligent urban waterlogging disposal method are implemented.

[0149] In addition, another embodiment of the present invention further provides a computer program product, which stores a computer program. When the computer program is executed by a processor, the steps of the above-mentioned intelligent urban waterlogging treatment method are implemented.

[0150] Those skilled in the art will appreciate that although some embodiments herein include certain features included in other embodiments but not other features, the combination of features from different embodiments is intended to be within the scope of the present invention and to form different embodiments. For example, any of the claimed embodiments may be used in any combination.

[0151] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. An intelligent method for handling urban waterlogging, characterized in that: The method comprises: Construct the feature vector and knowledge graph of the current waterlogging event based on the multimodal monitoring data of the current waterlogging event; Based on the feature vector similarity and knowledge graph structure similarity, historical cases similar to the current flooding incident are selected from the preset historical case data set to obtain a similar case set; Based on the preset treatment effect evaluation indicators, the treatment effects achieved by each historical case in the similar case set after adopting the corresponding response strategy are scored, and the recommended set of response strategies that match the current waterlogging incident is selected based on the scores; Constructing an urban waterlogging simulation model, wherein the urban waterlogging simulation model divides the current urban area into grids, assigns attribute information to each grid node, and configures the position of each drainage device in the urban waterlogging simulation model according to pipe network distribution data in a preset urban model; The highest-scoring response strategy in the recommended response strategy set is selected as the optimal response strategy. Based on the optimal response strategy and the location of each drainage device in the urban waterlogging simulation model, a dynamic deployment logic for drainage equipment is generated to obtain an urban waterlogging simulation deduction model. The predicted water depth at each grid node when a specified end time is reached is predicted based on the urban waterlogging simulation deduction model to obtain a simulation prediction result corresponding to the optimal response strategy. Conduct prevention and control measures in the current urban area based on the dynamic deployment logic of drainage equipment to obtain actual waterlogging monitoring results; The effectiveness of the current optimal response strategy is determined based on the simulation prediction results and the actual waterlogging monitoring results. If it is effective, the current optimal response strategy will continue to be implemented. Otherwise, the optimal response strategy will be reselected to deal with the current waterlogging event.

2. The method according to claim 1, characterized in that The feature vector of the current waterlogging event is constructed based on the multimodal monitoring data of the current waterlogging event, including: Obtain the current rainfall intensity, topography, pipe network flow, and water depth of the flooding event, and perform normalization and time alignment preprocessing on the acquired data; Extract feature data of the pre-processed precipitation intensity, terrain, pipe network flow and water depth to obtain precipitation intensity feature vector, terrain feature vector, pipe network flow feature vector and water depth feature vector; Based on the attention mechanism, the feature fusion of precipitation intensity feature vector, terrain feature vector, pipe network flow feature vector and water depth feature vector is performed, and the obtained fused feature vector is used as the feature vector of the current waterlogging event.

3. The method according to claim 2, characterized in that Based on the attention mechanism, feature fusion is performed on the precipitation intensity feature vector, terrain feature vector, pipe network flow feature vector, and water depth feature vector, including: The precipitation intensity feature vector, terrain feature vector, pipe network flow feature vector and water depth feature vector are concatenated into a long vector in the column direction; Construct a learnable weight matrix, perform a linear transformation on the concatenated long vector based on the weight matrix to obtain an attention score vector, and convert the attention score vector into an attention weight vector through the Softmax function to obtain the attention weights of the precipitation intensity feature vector, terrain feature vector, pipe network flow feature vector, and water depth feature vector when performing feature fusion; The precipitation intensity feature vector, terrain feature vector, pipe network flow feature vector and water depth feature vector are weightedly spliced ​​according to their corresponding attention weights to obtain a fused feature vector.

4. The method according to claim 1, wherein Based on the feature vector similarity and knowledge graph structure similarity, historical cases similar to the current flooding incident are selected from the preset historical case dataset, including: The text data in the historical case dataset is annotated with entities and relationships. The entities include disaster event entities, situation feature entities, disposal measure entities, and response result entities. The relationships represent the associations between different entities. Entities and relationships in the text data are extracted to form structured entity-relationship-entity triples. The structured triples are semantically stored in a knowledge graph to obtain a historical case knowledge graph. Construct a corresponding feature vector for each historical case in the historical case dataset; Calculate the similarity between the feature vectors of each historical case and the feature vector of the current waterlogging event; Extract the subgraph of each historical case from the historical case knowledge graph, and calculate the knowledge graph structure similarity between the subgraph of each historical case and the knowledge graph of the current flooding incident; The event similarity between each historical case and the current flooding event is calculated based on the feature vector similarity, knowledge graph structure similarity, and time decay factor. The event similarity calculation formula is as follows: ； in, is the event similarity, W, V, and U are the weight coefficients corresponding to the feature vector similarity, knowledge graph structure similarity, and time decay factor, respectively. is the characteristic vector of the current flooding event, is the feature vector of the historical case, is the L2 norm of the vector, is the knowledge graph of the current flooding event, which includes the entities and relationships of the current flooding event. is the subgraph of the i-th historical case, including the entities and relationships of the historical case, for and The knowledge graph structure similarity of is the preset time attenuation coefficient, The timestamp of the current flooding event. Timestamp for historical cases; Historical cases whose event similarity is greater than a preset similarity threshold are selected from a preset historical case data set to obtain historical cases similar to the current waterlogging event.

5. The method according to claim 1, wherein The treatment effect of each historical case in the similar case set after adopting the corresponding response strategy is scored according to the preset treatment effect evaluation indicators, including: Configure the weights of each treatment effect evaluation indicator; The scores of the treatment effects of each historical case after adopting the corresponding response strategy are calculated based on the preset scoring function. The scoring function is as follows: ； ， in, is the weight of the j-th treatment effect evaluation index, is the preset magnification factor, is the importance score of the jth treatment effect evaluation index, m is the total number of treatment effect evaluation indicators, g A temporary index for the sum operation. For the g The importance score of each treatment effect evaluation indicator, For strategy The final weighted total score of For strategy In the j The scores on the treatment effect evaluation indicators are: For the i candidate coping strategies.

6. The method according to claim 1, characterized in that Assign attribute information to each grid node and configure the location of each drainage device in the urban waterlogging simulation model based on the pipe network distribution data in the preset city model, including: According to the precipitation intensity, terrain and water depth data, each grid node in the urban waterlogging simulation model is configured with rainfall intensity, elevation value and water depth; According to the physical geometry information in the preset urban model, the center coordinates and area of ​​each grid node in the urban waterlogging simulation model are configured; The position of each drainage device in the urban waterlogging simulation model and the flow rate and influence radius of the corresponding drainage device are configured according to the pipe network flow and the pipe network distribution data in the city model.

7. The method according to claim 1, characterized in that The predicted value of the water depth at each grid node when the specified end time is reached is predicted according to the urban waterlogging simulation model, including: The urban waterlogging simulation model is used to predict the evolution of waterlogging to construct a continuous shallow water dynamics differential equation, which is as follows: ； ； ； in, is the change in water depth per unit time for each grid node, is the divergence operator, d is the water diffusivity, is the spatial gradient of water accumulation, h is the depth of water accumulation, For the grid nodes i The current optimal response strategy The net rate of change caused by is the number of drainage equipment, For the k The flow rate of drainage equipment, For the k The influence radius of each drainage device, For grid nodes i The center coordinates of For the k The deployment location coordinates of each drainage device, For net rain strength; is the measured rainfall intensity at position (x, y), Land use type L associated loss coefficients; The explicit finite difference method is used to discretize the continuous shallow water dynamics differential equation, and the discretized water depth prediction formula is obtained as follows: ； in, , in, For grid nodes i The depth of water at time step n+1 is, For grid nodes i At time step n The depth of water accumulation, is the time step, For grid nodes i The area, For grid nodes i The set of adjacent grid nodes, For grid nodes i and adjacent grid nodes j The diffusion rate at the interface, For Grid i and j The interface length between is the grid cell center distance, At the grid node i The net rate of change caused by the current optimal response strategy; When the urban waterlogging simulation model reaches the specified end time, the predicted value of the water depth of each grid node is obtained according to the water depth prediction formula.

8. The method according to claim 1, characterized in that The determination of whether the current optimal response strategy is effective based on the simulation prediction results and the actual waterlogging monitoring results includes: Calculate the absolute water depth error and relative error rate between the predicted water depth value of each grid node in the simulation prediction results when executing the current optimal response strategy and the water depth monitoring value of the actual monitoring point corresponding to each grid node; The effectiveness of the current optimal response strategy is judged based on the absolute water depth error and the relative error rate. If both the absolute water depth error and the relative error rate are less than the preset error tolerance threshold, the current optimal response strategy is determined to be effective.

9. The method according to claim 4, characterized in that The method further comprises: The current flooding incident and the corresponding optimal response strategy are used as new cases to be added, and entity extraction and relationship extraction are performed on the new cases to form candidate structured triples of entity-relationship-entity; When the similarity between the candidate entity extracted from the new case and all the existing entities in the historical case knowledge graph is less than the preset entity similarity judgment threshold, the candidate entity is added to the existing entity set to achieve incremental update of the knowledge base; When the similarity between the candidate entity extracted from the new case and all existing entities in the historical case knowledge graph is less than the preset entity similarity judgment threshold, and when the candidate structured triple is different from the existing triple relationship set in the historical case knowledge graph, the candidate structured triple is added to the existing relationship set to realize incremental update of the knowledge base.

10. An intelligent urban waterlogging treatment system, characterized in that: The system comprises: Multimodal data fusion module, used to construct the feature vector and knowledge graph of the current waterlogging event based on the multimodal monitoring data of the current waterlogging event; The graph reasoning module is used to select historical cases similar to the current flooding incident from a preset historical case dataset based on feature vector similarity and knowledge graph structure similarity to obtain a set of similar cases; The strategy recommendation module is used to score the treatment effects achieved by each historical case in the similar case set after adopting the corresponding response strategy based on the preset treatment effect evaluation index, and select the recommended set of response strategies that match the current waterlogging incident based on the score; A simulation model construction module is used to construct an urban waterlogging simulation model, which divides the current urban area into grids, assigns attribute information to each grid node, and configures the location of each drainage device in the urban waterlogging simulation model based on the pipe network distribution data in the preset urban model; A prediction and deduction module is used to select the response strategy with the highest score in the recommended response strategy set as the optimal response strategy, generate dynamic deployment logic for drainage equipment based on the optimal response strategy and the location of each drainage equipment in the urban waterlogging simulation model, obtain the urban waterlogging simulation and deduction model, and predict the water depth of each grid node at the specified end time based on the urban waterlogging simulation and deduction model to obtain the simulation prediction result corresponding to the optimal response strategy; The actual measurement acquisition module is used to carry out prevention and control measures in the current urban area based on the dynamic deployment logic of drainage equipment and obtain the actual waterlogging monitoring results; The strategy execution detection module is used to determine whether the current optimal response strategy is effective based on the simulation prediction results and the actual waterlogging monitoring results. If it is effective, the current optimal response strategy will continue to be executed; otherwise, the optimal response strategy will be reselected to deal with the current waterlogging event.

11. A computer device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and executable on the processor; when the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 9 are implemented.

12. A computer program product, characterized in that The computer program product stores a computer program, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 9 are implemented.

Images