Model establishment method and device for rail transit flood prevention, server and medium

Abstract

The invention discloses a model establishment method and device for rail transit flooding prevention, a server and a medium, and the method comprises the steps: obtaining digital twin data of a rail transit station, and determining the number of flooding prevention key points and corresponding flooding prevention attribute categories according to the digital twin data; when the total number of the anti-flooding input control key points, the anti-flooding transitive key points, the anti-flooding decisive key points and the anti-flooding control key points is greater than a first threshold value, establishing a dimensionality reduction improved feedforward neural network; and when the sum of the number of 1 / 4 of the anti-flooding input control key points, the number of the anti-flooding transitive key points, the number of the anti-flooding decisive key points and the number of the anti-flooding control key points is smaller than a first threshold value, establishing a GNN model. The complexity of the current site can be determined, and a proper model is selected according to the complexity. By means of the appropriate model, the calculation amount can be reduced, and meanwhile the over-fitting phenomenon generated due to the fact that the model is too complex can be avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of AI models, and in particular, to a method, device, server and medium for establishing a model for preventing flooding in rail transit. Background Art

[0002] Urban rail transit is an indispensable part of the current urban infrastructure layout in China. As an important component of urban sustainable development, it can effectively ease the contradiction between people and urban traffic congestion. With the global climate change and frequent extreme rainfall events, the flood prevention system has become an important part of urban resilience construction. Due to the particularity of urban rail transit, once waterlogging backflow occurs, it will cause the suspension of operation at least, affecting urban traffic and people's travel, and at worst, causing huge property losses. Therefore, during the preliminary construction and later renovation planning and design, it is necessary to effectively evaluate the flood prevention performance of rail transit stations, and then reasonably plan flood prevention equipment to ensure that the flood prevention performance of rail transit stations meets the flood prevention requirements.

[0003] In the prior art, a neural network model can be used to effectively evaluate the flood prevention performance of rail transit stations. By inputting the data of each station and the collected water accumulation situation, the inundation situation of each key part in the station can be obtained. In the process of implementing the present invention, the inventor found the following technical problems: due to the existence of various types of stations such as elevated, ordinary, and complex transfer stations in rail transit stations, and the data types of stations are diverse and complex. If the AI model is relatively simple, although it can produce good prediction results for elevated and ordinary stations, for complex transfer stations, the accuracy will be greatly reduced. If the AI model is relatively complex, a large amount of missing data will be generated for elevated and ordinary stations, and overfitting is likely to occur, and the prediction accuracy for this scenario will be greatly reduced. Summary of the Invention

[0004] The purpose of the present invention is to overcome the deficiencies in the prior art and solve the technical problem of low accuracy when using a single model to evaluate and predict the flood prevention performance of rail transit stations in the prior art, and to provide a method, device, server and medium for establishing a model for preventing flooding in rail transit.

[0005] The purpose of the present invention is achieved through the following technical solutions:

[0006] A method for establishing a model for preventing flooding in rail transit, comprising:

[0007] Obtain the digital twin data of the rail transit station, and determine the number of flood prevention key points and the corresponding flood prevention attribute categories according to the digital twin data, where the flood prevention attribute categories include flood prevention input control key points, flood prevention transitivity key points, flood prevention decisive key points and flood prevention control key points;

[0008] When the total number of flood prevention input control key points, flood prevention transitive key points, flood prevention decisive key points, and flood prevention control key points is greater than the first threshold, a dimensionality reduction improved feedforward neural network is established. The dimensionality reduction improved feedforward neural network takes the data of the flood prevention input control key points and flood prevention control key points as inputs and outputs the predicted data of the flood prevention transitive key points and flood prevention decisive key points.

[0009] When the sum of the number of one-fourth of the flood prevention input control key points, flood prevention transitive key points, flood prevention decisive key points, and flood prevention control key points is less than the first threshold, a GNN model is established. The GNN model establishes and updates the connection relationships between hierarchical nodes based on the clustering nodes of the flood prevention input control key points, the hierarchical relationships of the flood prevention transitive key points, flood prevention decisive key points, and flood prevention control key points, continuously updates each hierarchical node according to the data of the flood prevention input control key points, and obtains the predicted classification of the flood prevention transitive key points and flood prevention decisive key points based on the updated hierarchical nodes.

[0010] Further, the GNN model includes:

[0011] An aggregation layer for establishing an input node graph based on the positional relationship of the flood prevention input control key points and generating flood prevention input aggregation nodes using an aggregation function.

[0012] A graph establishment layer for generating a directed graph using the positional and water flow sequence relationships of the flood prevention input aggregation nodes, flood prevention transitive key points, flood prevention decisive key points, and flood prevention control key points, where the flood prevention control key points are negative.

[0013] A fully connected layer for outputting the flood prevention possibility probabilities of the flood prevention transitive key points and flood prevention decisive key points using the attributes of the flood prevention transitive key points and flood prevention decisive key points output by the graph establishment layer.

[0014] Further, the acquisition of the digital twin data of the rail transit station includes:

[0015] Obtaining virtual precipitation data predicted by a meteorological large model;

[0016] Determining the corresponding flood prevention management strategy according to the virtual precipitation data and updating the digital twin data based on the flood prevention management strategy.

[0017] Further, the determination of the corresponding flood prevention management strategy according to the virtual precipitation data and the update of the digital twin data based on the flood prevention management strategy include:

[0018] Adjusting the height of the water retaining facilities and the drainage power of the peripheral drainage facilities according to the precipitation data and the operation comparison table of the water retaining facilities and peripheral drainage facilities;

[0019] Update the digital twin data according to the adjusted water retaining facilities and peripheral drainage facilities.

[0020] Further, determining the number of flood prevention key points and the corresponding flood prevention attribute categories according to the digital twin data includes:

[0021] Determine each flood prevention attribute category according to the flood prevention requirements, and select a number of flood prevention points according to the digital twin data according to the flood prevention transfer key points, flood prevention decisive key points and flood prevention control key points;

[0022] Determine the redundancy index and hot standby index of each flood prevention point according to a number of flood prevention points;

[0023] Calculate the vulnerability of each flood prevention point according to the redundancy index and hot standby index, and calculate the stability of each flood prevention point;

[0024] Select flood prevention key points from the flood prevention points according to the vulnerability and stability.

[0025] Further, selecting a number of flood prevention points according to each category according to the digital twin data includes:

[0026] Divide the stairs in the site according to the grid unit according to the site digital twin data;

[0027] Select grid units from the bottom of the stairs, the side walls of the middle floors, and the middle floor positions as flood prevention transfer key points;

[0028] Use the turbulence model to calculate the water flow velocity of each flood prevention transfer key point according to the data of the flood prevention input control key point.

[0029] Further, the dimension reduction improved feedforward neural network includes:

[0030] An input layer for converting the location, predicted precipitation, time of predicted precipitation, spatial position data, redundancy index and hot standby index of the flood prevention input control key point and the flood prevention control key point into multi-dimensional data of flood prevention points;

[0031] A dimension reduction layer for performing dimension reduction processing on the multi-dimensional data of flood prevention points to obtain the dimension reduction data of the multi-dimensional data of flood prevention points;

[0032] Multiple hidden layers for extracting abstract features through linear transformation;

[0033] Two output layers, where one output layer is used to output the flood prevention prediction result using the activation function according to the abstract features, and the other output layer is used to output the flow velocity prediction result of the flood prevention transfer control point according to the abstract features output by some hidden layers;

[0034] An adjustment unit for adjusting the bias matrix corresponding to the partial hidden layer according to the result calculated by the turbulence model.

[0035] The present invention also provides a model establishment device for rail transit flood prevention, including:

[0036] An acquisition module for acquiring digital twin data of a rail transit station, determining the number of flood prevention key points and corresponding flood prevention attribute categories according to the digital twin data, where the flood prevention attribute categories include flood prevention input control key points, flood prevention transitivity key points, flood prevention decisive key points, and flood prevention control key points;

[0037] A first establishment module for establishing a dimensionality reduction improved feedforward neural network when the total number of flood prevention input control key points, flood prevention transitivity key points, flood prevention decisive key points, and flood prevention control key points is greater than a first threshold. The dimensionality reduction improved feedforward neural network takes the data of the flood prevention input control key points and flood prevention control key points as inputs and outputs prediction data of flood prevention transitivity key points and flood prevention decisive key points;

[0038] A second establishment module for establishing a GNN model when the sum of the number of one-quarter of the flood prevention input control key points, flood prevention transitivity key points, flood prevention decisive key points, and flood prevention control key points is less than the first threshold. The GNN model establishes and updates the connection relationships between hierarchical nodes according to the clustering nodes of the flood prevention input control key points, the hierarchical relationships of the flood prevention transitivity key points, flood prevention decisive key points, and flood prevention control key points, continuously updates each hierarchical node according to the data of the flood prevention input control key points, and obtains the prediction classification of the flood prevention transitivity key points and flood prevention decisive key points according to the updated hierarchical nodes.

[0039] The present invention also provides a server, where the server includes:

[0040] One or more processors;

[0041] A storage device for storing one or more programs;

[0042] When the one or more programs are executed by the one or more processors, the one or more processors implement the model establishment method for rail transit flood prevention described above.

[0043] The present invention also provides a storage medium containing computer-executable instructions, where the computer-executable instructions are used to execute the model establishment method for rail transit flood prevention described above when executed by a computer processor.

[0044] Compared with the prior art, the beneficial effects brought by the technical solution of the present invention are:

[0045] 1. The present invention can accurately capture the actual situation of the station and judge the complexity: By obtaining the digital twin data of the rail transit station, the information such as the physical space, meteorological data, and flood prevention measures is digitalized and virtualized, enabling flood prevention assessment to be based on real and dynamic data. This data-driven mode can reflect the structure and operation status of the station in real time.

[0046] Using the digital twin data, multiple flood prevention points are set within the station according to flood prevention requirements. Then, through calculations of redundancy, hot standby, vulnerability, and stability, the key points are accurately identified from multiple potential flood prevention points and classified into categories such as "flood prevention input control", "flood prevention transitivity", "flood prevention determination", and "flood prevention control", thereby capturing the influence of different key factors on the flood prevention effect. This multi-dimensional data extraction and analysis effectively reduces the risk of missing key factors and provides comprehensive and detailed data support for subsequent model establishment.

[0047] 2. The present invention can achieve flexible selection of models: According to the total number of key points within the station and the proportional relationship between various categories (for example, when the number of flood prevention input control key points is sufficient and the total number is higher than the first threshold, a dimensionality reduction improved feedforward neural network is adopted; when the number of key points is small and the proportion is lower than the threshold, a graph neural network GNN is adopted), adaptive model selection is realized.

[0048] After the dimensionality reduction improved feedforward neural network inputs the data of flood prevention input control key points and flood prevention control key points, it eliminates data redundancy through the dimensionality reduction layer, extracts abstract features through multiple hidden layers, and finally outputs the prediction results for flood prevention transitivity and flood prevention determination key points. This design not only reduces the data dimension and computational complexity but also avoids the overfitting problem caused by an overly complex model.

[0049] For the complex station structure and small amount of data scenario, the graph neural network (GNN) model uses the GNN model to establish the connection relationship between each level of nodes (including aggregation nodes, update level nodes, etc.), and conducts information transmission, fusion, and classification according to natural laws such as spatial position and water flow direction to ensure the classification accuracy of the prediction. This model is particularly suitable for flood prevention risk assessment in heterogeneous, multi-structured, and few-sample situations.

[0050] 3. The present invention combines virtual precipitation data and flood prevention management strategies: The present invention uses a meteorological large model to predict and obtain virtual precipitation data, and dynamically adjusts the physical facility parameters (such as the height of the water retaining facility and the drainage power of the drainage facility) according to the operation comparison table of precipitation data and water retaining facilities and peripheral drainage facilities, so that the digital twin data is updated in real time.

[0051] Moreover, based on the closed-loop feedback mechanism, the cycle update of the digital twin data based on the actual rainfall prediction data and the adjusted facilities not only enables the model prediction to have a real-time correction function, but also provides timely feedback and improvement basis for the flood prevention management strategy, thereby effectively preventing risks under extreme rainfall events.

[0052] 4. The present invention can reduce the operation burden and improve the prediction accuracy: By accurately judging the complexity of the site data, the present invention adaptively selects a simplified or complex model, which not only effectively reduces the calculation amount, but also avoids the overfitting problem caused by too little data, and finally achieves higher prediction accuracy.

[0053] The input data is preprocessed by the dimensionality reduction layer to remove redundant information, ensuring that the subsequent hidden layer can more efficiently extract key abstract features, effectively improving the accuracy and robustness of the prediction results.

[0054] 5. The present invention provides a scientific basis for flood prevention measure planning: The key point data acquisition, key model selection and real-time closed-loop update of the present invention make the site flood prevention performance evaluation more accurate, which provides a quantitative and verifiable scientific basis for the planning and improvement of urban rail transit flood prevention measures, thereby facilitating the rational allocation of flood control resources, reducing accident risks, and mitigating the economic losses and social impacts brought by extreme rainfall events.

[0055] In addition, different evaluation strategies are adopted for different types of sites such as elevated, ordinary and complex transfer stations. The present invention provides a technical solution with wide adaptability and refined risk management, which can effectively predict flood prevention risks in different scenarios and achieve the improvement of overall benefits.

[0056] In summary, through the real-time acquisition and refined processing of digital twin data, multi-angle classification of key points, model adaptive selection and dynamic strategy regulation, the present invention can significantly improve the accuracy of flood prevention performance evaluation of rail transit stations, reduce the model calculation complexity and risk overfitting, thereby providing a reliable, effective and scientific decision-making basis for the reasonable planning of flood prevention measures and urban traffic safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 is a schematic flowchart of a method for establishing a model for rail transit flood prevention provided in Embodiment 1 of the present invention;

[0058] Figure 2 is a schematic flowchart of a method for establishing a model for rail transit flood prevention provided in Embodiment 2 of the present invention;

[0059] Figure 3 is a schematic structural diagram of a device for establishing a model for rail transit flood prevention provided in Embodiment 3 of the present invention;

[0060] Figure 4 It is a schematic structural diagram of a server provided in Embodiment 4 of the present invention. Detailed implementation manners

[0061] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only for explaining the present invention, rather than limiting the present invention. In addition, it should be noted that, for the sake of convenience of description, only parts related to the present invention are shown in the drawings, rather than all the structures.

[0062] Embodiment 1

[0063] Figure 1 It is a flowchart of a method for establishing a model for preventing flooding in rail transit provided in Embodiment 1 of the present invention. This embodiment is applicable to the situation of establishing a model for preventing flooding in rail transit, especially for preventing flooding at stations. This method can be executed by a device for establishing a model for preventing flooding in rail transit, and specifically includes the following steps:

[0064] Step 110, obtain digital twin data of a rail transit station, and determine the number of flood prevention key points and corresponding flood prevention attribute categories according to the digital twin data. The flood prevention attribute categories include flood prevention input control key points, flood prevention transfer key points, flood prevention decisive key points, and flood prevention control key points.

[0065] When planning a rail transit station, generally, a high-precision three-dimensional model can be constructed based on Building Information Modeling (BIM) technology to truly reproduce the station space structure and equipment distribution, such as turnstiles, elevators, ventilation systems, etc. Digital twin data of the rail transit station can be obtained from the BIM, and digital twin data can also be obtained according to the data collected by current actual sensors.

[0066] In this embodiment, the flood prevention key points can be divided into flood prevention input control key points, flood prevention transfer key points, flood prevention decisive key points, and flood prevention control key points. The flood prevention input control key point can be a point where external rainwater enters the station. Exemplarily, it can be a key point corresponding to an entrance, a construction interface, etc. The flood prevention transfer key point can be a point where various risks may occur during the inflow of external rainwater from the flood prevention input control key point. The flood prevention decisive key point can be a key point where risks will occur once flooded, such as power supply, communication, environmental control, elevators, and ticketing systems. The flood prevention control key point can be a point where various drainage devices are located, and it can be a key point that can control the flooding degree to a certain extent.

[0067] Further, the acquisition of the digital twin data of the rail transit station may include: acquiring virtual precipitation data predicted by a meteorological large model; determining a corresponding flood prevention management strategy based on the virtual precipitation data, and updating the digital twin data based on the flood prevention management strategy. The determining a corresponding flood prevention management strategy based on the virtual precipitation data and updating the digital twin data based on the flood prevention management strategy may include: adjusting the height of the water retaining facilities and the drainage power of the peripheral drainage facilities according to the precipitation data and the operation control table of the water retaining facilities and the peripheral drainage facilities; updating the digital twin data according to the adjusted water retaining facilities and peripheral drainage facilities. Usually, the station can adopt water retaining facilities and peripheral drainage facilities according to the water accumulation situation corresponding to the predicted rainfall intensity to reduce the water accumulation flowing into the station. In this case, the rainwater flowing into the station will also change. Therefore, the digital twin data can be updated according to the adjusted water retaining facilities and peripheral drainage facilities to better simulate the flooding situation.

[0068] Further, since there are multiple devices corresponding to the power supply, communication, environmental control, elevator and ticketing systems in the station respectively, if each device is set as a key point, it will not only increase the data volume, but also increase the calculation amount. Therefore, in this embodiment, representative ones can be selected from the above devices as flood prevention key points.

[0069] Exemplarily, it may include: determining each flood prevention attribute category according to the flood prevention requirements, and selecting multiple flood prevention points according to the digital twin data as flood prevention transfer key points, flood prevention decisive key points and flood prevention control key points; determining the redundancy index and hot standby index of each flood prevention point according to the multiple flood prevention points; calculating the vulnerability of each flood prevention point according to the redundancy index and the hot standby index, and calculating the stability of each flood prevention point; selecting flood prevention key points from the flood prevention points according to the vulnerability and stability. By using the above method, the station can be set in the lowest reliability state to determine the devices that can operate normally and use them as flood prevention key points.

[0070] Step 120, when the total number of the flood prevention input control key points, flood prevention transfer key points, flood prevention decisive key points and flood prevention control key points is greater than the first threshold, establish a dimensionality reduction improved feedforward neural network, and the dimensionality reduction improved feedforward neural network takes the data of the flood prevention input control key points and the flood prevention control key points as inputs and outputs the predicted data of the flood prevention transfer key points and the flood prevention decisive key points.

[0071] In this embodiment, the first threshold can be set according to the experience value based on the performance of the server and the processing efficiency of the model. Calculate the total number of flood prevention input control key points, flood prevention transitive key points, flood prevention decisive key points, and flood prevention control key points. When the total number exceeds the first threshold, it can be determined that the current site is a complex site with a large amount of data to be processed. Therefore, a more complex model needs to be adopted to obtain better processing results.

[0072] Optionally, the model can adopt a dimensionality reduction improved feedforward neural network. Through dimensionality reduction processing, complex data can be reduced in dimension, and irrelevant data dimensions can be reduced. The amount of compression operation is reduced. At the same time, the rainwater flow is a linear relationship. Using the advantage of the linear processing of the feedforward neural network, better processing results can be obtained for complex sites. The dimensionality reduction improved feedforward neural network takes the data of the flood prevention input control key points and the flood prevention control key points as inputs and outputs the predicted data of the flood prevention transitive key points and the flood prevention decisive key points.

[0073] Step 130, when the sum of the number of flood prevention input control key points that is one-fourth, flood prevention transitive key points, flood prevention decisive key points, and flood prevention control key points is less than the first threshold, establish a GNN model. The GNN model establishes and updates the connection relationship between hierarchical nodes according to the hierarchical relationship of the clustering nodes, flood prevention transitive key points, flood prevention decisive key points, and flood prevention control key points of the flood prevention input control key points, continuously updates each hierarchical node according to the data of the flood prevention input control key points, and obtains the predicted classification of the flood prevention transitive key points and the flood prevention decisive key points according to the updated hierarchical nodes.

[0074] In this embodiment, if the number of key points is small, it can be considered an ordinary site. Based on this situation, a GNN model can be selected. The graph neural network (GNN) is a deep learning model based on graph-structured data, which realizes feature learning and reasoning by aggregating node, edge, and their neighborhood information. Information is transmitted and features are updated through the graph structure (the topological relationship between nodes and edges). It is particularly suitable for scenarios of water flow.

[0075] Although the GNN model performs excellently in the scenario provided in this embodiment, too many node numbers will seriously affect the computing efficiency and consume more computing resources. Therefore, in this embodiment, the number of nodes, that is, the number of key points, needs to be considered first.

[0076] Exemplarily, the number of a quarter of the flood prevention input control key points, the flood prevention transitive key points, the flood prevention decisive key points, and the sum of the number of flood prevention control key points can be calculated, and then the sum is compared with a first threshold. When it is less than the first threshold, it can be determined that the number of key points is suitable for using the GNN model. The reason for considering the number of a quarter of the flood prevention input control key points is that the flood prevention input control key points can compress the flood prevention input control key points in an aggregation manner, but excessive aggregation may affect the prediction accuracy. Therefore, in this embodiment, a hierarchical clustering method can be adopted to perform clustering twice, and the clustered flood prevention input control key points are used as node inputs to the GNN model. If the number of flood prevention input control key points is odd, the number of nodes output by hierarchical clustering can be used as a constraint to ensure that the number of nodes after clustering is the number of a quarter of the flood prevention input control key points. The GNN model establishes and updates the connection relationship between hierarchical nodes according to the hierarchical relationship between the clustered nodes of the flood prevention input control key points, the flood prevention transitive key points, the flood prevention decisive key points, and the flood prevention control key points, continuously updates each hierarchical node according to the data of the flood prevention input control key points, and obtains the predicted classification of the flood prevention transitive key points and the flood prevention decisive key points based on the updated hierarchical nodes.

[0077] In this embodiment, the GNN model includes: a combination layer for establishing an input node graph according to the positional relationship of the flood prevention input control key points and generating flood prevention input aggregation nodes by using an aggregation function; a graph establishment layer for generating a directed graph by using the positional and water flow sequence relationships of the flood prevention input aggregation nodes, the flood prevention transitive key points, the flood prevention decisive key points, and the flood prevention control key points, wherein the flood prevention control key point is negative; and a fully connected layer for outputting the flood prevention possibility probabilities of the flood prevention transitive key points and the flood prevention decisive key points by using the attributes of the flood prevention transitive key points and the flood prevention decisive key points output by the graph establishment layer.

[0078] Exemplarily, key points, as well as various data of the aggregated key points, such as position, predicted precipitation, spatial position of the location, etc., can be used as attributes of the nodes. By iterating the relationship between the nodes and edges in the graph, and then using the fully connected layer, the attributes of the flood prevention transitive key points and the flood prevention decisive key points are obtained, and the flood prevention possibility probabilities of the flood prevention transitive key points and the flood prevention decisive key points are output based on the attributes. Since the flood prevention control key points can be various drainage devices, and in essence, it is an inverse flow process of water flow, it can be set to negative.

[0079] In this embodiment, by obtaining the digital twin data of the rail transit station, the number of flood prevention key points and the corresponding flood prevention attribute categories are determined according to the digital twin data. The flood prevention attribute categories include flood prevention input control key points, flood prevention transmissibility key points, flood prevention decisive key points, and flood prevention control key points. When the total number of flood prevention input control key points, flood prevention transmissibility key points, flood prevention decisive key points, and flood prevention control key points is greater than the first threshold, a dimensionality reduction improved feedforward neural network is established. The dimensionality reduction improved feedforward neural network takes the data of the flood prevention input control key points and the flood prevention control key points as inputs and outputs the predicted data of the flood prevention transmissibility key points and the flood prevention decisive key points. When the sum of the number of one-fourth of the flood prevention input control key points, the flood prevention transmissibility key points, the flood prevention decisive key points, and the flood prevention control key points is less than the first threshold, a GNN model is established. The GNN model establishes and updates the connection relationships between the hierarchical nodes according to the clustering nodes of the flood prevention input control key points, the hierarchical relationships of the flood prevention transmissibility key points, the flood prevention decisive key points, and the flood prevention control key points, continuously updates each hierarchical node according to the data of the flood prevention input control key points, and obtains the predicted classification of the flood prevention transmissibility key points and the flood prevention decisive key points based on the updated hierarchical nodes. By using the digital twin data of the rail transit station, various key points and corresponding types required by the model are determined, and the number of key points of each type is used to further improve the accuracy of the station flooding prediction, thereby providing an accurate reference basis for the planning and improvement of flood prevention measures.

[0080] Embodiment 2

[0081] Figure 2It is a schematic flowchart of the model establishment method for rail transit flood prevention provided in the second embodiment of the present invention. This embodiment is optimized based on the above embodiment, and the dimensionality reduction improved feedforward neural network is specifically optimized as follows: an input layer, which is used to convert the flood prevention input control key points, the locations of the flood prevention control key points, the predicted precipitation, the time of the predicted precipitation, the spatial position data, the redundancy index, and the hot standby index into multi-dimensional flood prevention point data; a dimensionality reduction layer, which is used to perform dimensionality reduction processing on the multi-dimensional data to obtain the dimensionality reduction data of the multi-dimensional data; multiple hidden layers, which are used to extract abstract features through linear transformation; two output layers, where one output layer is used to output the flood prevention prediction result using an activation function based on the abstract features, and the other output layer is used to output the flow velocity prediction result of the flood prevention transfer control points based on the abstract features output by some of the hidden layers; an adjustment unit, which is used to adjust the bias matrix corresponding to the partial hidden layer according to the result calculated by the turbulence model. Correspondingly, the step of selecting multiple flood prevention points according to each category from the digital twin data is specifically optimized as follows: dividing the stairs in the station according to grid cells based on the station digital twin data; selecting grid cells from the bottom of the stairs, the side walls of the middle layer, and the middle layer position as the flood prevention transfer key points; using the turbulence model to calculate the water flow velocity of each flood prevention transfer key point according to the data of the flood prevention input control key points.

[0082] See Figure 2 , the model establishment method for rail transit flood prevention includes:

[0083] Step 210, obtain the digital twin data of the rail transit station, divide the stairs in the station according to grid cells based on the station digital twin data; select grid cells from the bottom of the stairs, the side walls of the middle layer, and the middle layer position as the flood prevention transfer key points.

[0084] In a rail transit station, the stairs are the main channels for people to go up and down. And the stairs have a certain height difference. When rainwater enters the entrance from the outside, it will quickly flow into the lower layer of the station through the stairs. If the flow velocity is too fast, it will pose a greater risk. Therefore, in this embodiment, the stairs in the station can be divided according to grid cells based on the station digital twin data; select grid cells from the bottom of the stairs, the side walls of the middle layer, and the middle layer position as the flood prevention transfer key points. The flow velocity at the bottom of the stairs is the fastest, and due to the impact of the side walls, a relatively large flow velocity impact will be generated on the side walls of the middle layer. Since there is a stair buffer flat layer in the middle layer, the flow velocity at the middle layer position is also required. By dividing it into multiple units and selecting some points as the flood prevention transfer key points.

[0085] Step 220, use the turbulence model to calculate the water flow velocity of each flood prevention transfer key point according to the data of the flood prevention input control key points.

[0086] In this embodiment, the turbulence model can adopt the k-ε model, and the water flow velocity of each flood prevention transfer key point is calculated by using the input rainwater data of the flood prevention input control key points, especially the key points corresponding to the stairs.

[0087] Step 230, when the total number of flood prevention input control key points, flood prevention transfer key points, flood prevention decisive key points, and flood prevention control key points is greater than the first threshold, a dimensionality reduction improved feedforward neural network is established. The dimensionality reduction improved feedforward neural network includes: an input layer, which is used to convert the locations of flood prevention input control key points and flood prevention control key points, predicted precipitation, time of predicted precipitation, spatial position data, redundancy index, and hot standby index into multi-dimensional flood prevention point data; a dimensionality reduction layer, which is used to perform dimensionality reduction processing on the multi-dimensional data to obtain the dimensionality reduction data of the multi-dimensional data; multiple hidden layers, which are used to extract abstract features through linear transformation; two output layers, where one output layer is used to output the flood prevention prediction result by using the activation function according to the abstract features, and the other output layer is used to output the flow velocity prediction result of the flood prevention transfer control point according to the abstract features output by some of the hidden layers; an adjustment unit, which is used to adjust the bias matrix corresponding to the part of the hidden layers according to the result calculated by the turbulence model.

[0088] In this embodiment, in the dimensionality reduction improved feedforward neural network, the input layer can be used to convert various parameters such as the locations of flood prevention input control key points and flood prevention control key points, predicted precipitation, time of predicted precipitation, spatial position data, redundancy index, and hot standby index into multi-dimensional flood prevention point data.

[0089] Since the dimension of the input data attributes is relatively large, it is not convenient for later calculations. And due to the high dimension, multiple layers need to be set to extract features, which not only requires a large amount of computing power but also is prone to overfitting. Therefore, in this embodiment, a dimensionality reduction layer can be added. The dimensionality reduction layer can perform dimensionality reduction processing on the multi-dimensional data to obtain the dimensionality reduction data of the multi-dimensional data. Exemplarily, factor analysis can be used to perform dimensionality reduction processing on the multi-dimensional data.

[0090] The data after dimensionality reduction processing can be input into the hidden layer, and abstract features can be extracted through linear transformation, and then the flood prevention prediction results can be output using the activation function. In this embodiment, another output layer can also be provided, which uses an activation function to output the flow velocity prediction results of the flood prevention transfer control points for the abstract features output by part of the hidden layer. In the above steps, the flow velocity of the flood transfer key points can be calculated using the turbulence model. Using the comparison result between the two, the abstract features obtained by the hidden layer can be corrected. Since the hidden layer uses the method of linear transformation and there is a bias matrix for each hidden layer, different results can be obtained using the two, and the bias matrix can be optimized and adjusted to further improve the prediction results of the dimensionality reduction improved feedforward neural network.

[0091] Step 240, the dimensionality reduction improved feedforward neural network uses the data of the flood prevention input control key points and the flood prevention control key points as inputs, and outputs the prediction data of the flood prevention transfer key points and the flood prevention decisive key points.

[0092] Step 250, when the sum of the number of one-quarter of the flood prevention input control key points, the flood prevention transfer key points, the flood prevention decisive key points, and the flood prevention control key points is less than the first threshold, a GNN model is established. The GNN model establishes and updates the connection relationship between the hierarchical nodes according to the hierarchical relationship of the clustering nodes of the flood prevention input control key points, the flood prevention transfer key points, the flood prevention decisive key points, and the flood prevention control key points, continuously updates each hierarchical node according to the data of the flood prevention input control key points, and obtains the prediction classification of the flood prevention transfer key points and the flood prevention decisive key points according to the updated hierarchical nodes.

[0093] In this embodiment, the dimensionality reduction improved feedforward neural network is specifically optimized as follows: an input layer, which is used to convert the anti-flood input control key points, the locations of anti-flood control key points, predicted precipitation, the time of predicted precipitation, spatial position data, redundancy index, and hot standby index into multi-dimensional data of anti-flood points; a dimensionality reduction layer, which is used to perform dimensionality reduction processing on the multi-dimensional data to obtain the dimensionality-reduced data of the multi-dimensional data; multiple hidden layers, which are used to extract abstract features through linear transformation; two output layers, where one output layer is used to output the anti-flood prediction result using an activation function based on the abstract features, and the other output layer is used to output the flow velocity prediction result of the anti-flood transfer control point based on the abstract features output by some of the hidden layers; an adjustment unit, which is used to adjust the bias matrix corresponding to the part of the hidden layers according to the result calculated by the turbulence model. Correspondingly, the selection of multiple anti-flood points according to each category from the digital twin data is specifically optimized as follows: dividing the stairs in the site into grid cells according to the site digital twin data; selecting grid cells from the bottom of the stairs, the side walls of the middle layer, and the middle layer position as the anti-flood transfer key points; using the turbulence model to calculate the water flow velocity of each anti-flood transfer key point according to the data of the anti-flood input control key points. Through the dimensionality reduction characteristics of the dimensionality reduction improved feedforward neural network, a large amount of data can be effectively compressed, effectively reducing the complexity of the model. At the same time, the bias matrix of the hidden layer can be optimized and adjusted according to the calculation result of the flow velocity of the anti-flood transfer control point calculated by the turbulence model, reducing the problem of reduced calculation accuracy caused by dimensionality reduction.

[0094] Embodiment III

[0095] Figure 3 is a schematic structural diagram of a model establishment device for rail transit anti-flood provided by Embodiment III of the present invention. Refer to Figure 3 The model establishment device for rail transit anti-flood includes:

[0096] An acquisition module 310, which is used to acquire the digital twin data of the rail transit site, and determine the number of anti-flood key points and the corresponding anti-flood attribute categories according to the digital twin data. The anti-flood attribute categories include anti-flood input control key points, anti-flood transfer key points, anti-flood decisive key points, and anti-flood control key points;

[0097] A first establishment module 320, which is used to establish a dimensionality reduction improved feedforward neural network when the total number of anti-flood input control key points, anti-flood transfer key points, anti-flood decisive key points, and anti-flood control key points is greater than a first threshold. The dimensionality reduction improved feedforward neural network takes the data of the anti-flood input control key points and the anti-flood control key points as inputs and outputs the prediction data of the anti-flood transfer key points and the anti-flood decisive key points;

[0098] A second establishment module 330, configured to establish a GNN model when the sum of the number of a quarter of the flood prevention input control key points, the flood prevention transitive key points, the flood prevention decisive key points, and the flood prevention control key points is less than a first threshold. The GNN model establishes and updates the connection relationships between hierarchical nodes according to the hierarchical relationships of the clustering nodes of the flood prevention input control key points, the flood prevention transitive key points, the flood prevention decisive key points, and the flood prevention control key points, continuously updates each hierarchical node according to the data of the flood prevention input control key points, and obtains the predicted classifications of the flood prevention transitive key points and the flood prevention decisive key points based on the updated hierarchical nodes.

[0099] The model establishment device for rail transit flood prevention provided in this embodiment obtains the digital twin data of a rail transit station, determines the number of flood prevention key points and the corresponding flood prevention attribute categories according to the digital twin data. The flood prevention attribute categories include flood prevention input control key points, flood prevention transitive key points, flood prevention decisive key points, and flood prevention control key points. When the total number of the flood prevention input control key points, the flood prevention transitive key points, the flood prevention decisive key points, and the flood prevention control key points is greater than a first threshold, a dimensionality reduction improved feedforward neural network is established. The dimensionality reduction improved feedforward neural network takes the data of the flood prevention input control key points and the flood prevention control key points as inputs and outputs the predicted data of the flood prevention transitive key points and the flood prevention decisive key points. When the sum of the number of a quarter of the flood prevention input control key points, the flood prevention transitive key points, the flood prevention decisive key points, and the flood prevention control key points is less than a first threshold, a GNN model is established. The GNN model establishes and updates the connection relationships between hierarchical nodes according to the hierarchical relationships of the clustering nodes of the flood prevention input control key points, the flood prevention transitive key points, the flood prevention decisive key points, and the flood prevention control key points, continuously updates each hierarchical node according to the data of the flood prevention input control key points, and obtains the predicted classifications of the flood prevention transitive key points and the flood prevention decisive key points based on the updated hierarchical nodes. By using the digital twin data of the rail transit station, various key points and corresponding types required by the model are determined, the complexity of the current station is determined by using the number of key points of each type, and a suitable model is selected according to the complexity. Using a suitable model can not only reduce the amount of calculation, but also avoid the overfitting phenomenon caused by the model being too complex, further improve the accuracy of station flooding prediction, and thus provide an accurate reference basis for the planning and improvement of flood prevention measures.

[0100] Based on the above embodiments, the GNN model includes:

[0101] An aggregation layer, configured to establish an input node graph according to the positional relationships of the flood prevention input control key points, and generate flood prevention input aggregation nodes by using an aggregation function;

[0102] Graph building layer, which is used to generate a directed graph by using the positions and water flow sequence relationships of the flood prevention input aggregation nodes, flood prevention transitive key points, flood prevention decisive key points, and flood prevention control key points, wherein the flood prevention control key points are negative;

[0103] Fully connected layer, which is used to output the flood prevention possibility probabilities of the flood prevention transitive key points and flood prevention decisive key points by using the attributes of the flood prevention transitive key points and flood prevention decisive key points output by the graph building layer.

[0104] Based on the above embodiments, the acquisition module includes:

[0105] Acquisition unit, which is used to acquire virtual precipitation data predicted by the meteorological large model;

[0106] Update unit, which is used to determine the corresponding flood prevention management strategy according to the virtual precipitation data and update the digital twin data based on the flood prevention management strategy.

[0107] Based on the above embodiments, the update unit includes:

[0108] Adjustment subunit, which is used to adjust the height of the water retaining facilities and the drainage power of the peripheral drainage facilities according to the precipitation data and the operation comparison table of the water retaining facilities and peripheral drainage facilities;

[0109] Update subunit, which is used to update the digital twin data according to the adjusted water retaining facilities and peripheral drainage facilities.

[0110] Based on the above embodiments, the acquisition module includes:

[0111] Selection unit, which is used to determine each flood prevention attribute category according to the flood prevention requirements and select multiple flood prevention points according to the digital twin data according to the flood prevention transitive key points, flood prevention decisive key points, and flood prevention control key points;

[0112] Determination unit, which is used to determine the redundancy index and hot standby index of each flood prevention point according to multiple flood prevention points;

[0113] Calculation unit, which is used to calculate the vulnerability of each flood prevention point according to the redundancy index and hot standby index, and calculate the stability of each flood prevention point;

[0114] Flood prevention key point selection unit, which is used to select flood prevention key points from the flood prevention points according to the vulnerability and stability.

[0115] Based on the above embodiments, the selection unit includes:

[0116] Division subunit, which is used to divide the stairs in the site according to the grid unit according to the site digital twin data;

[0117] A selection unit for selecting grid cells as flood prevention transfer key points from the bottom of the stairs, the side walls of the middle floors, and the middle floor positions.

[0118] A speed calculation unit for calculating the water flow speed of each flood prevention transfer key point using a turbulence model based on the data of the flood prevention input control key points.

[0119] Based on the above embodiments, the dimensionality reduction improved feedforward neural network includes:

[0120] An input layer for converting the flood prevention input control key points, the locations of the flood prevention control key points, the predicted precipitation, the time of the predicted precipitation, the spatial position data, the redundancy index, and the hot standby index into flood prevention point multi-dimensional data.

[0121] A dimensionality reduction layer for performing dimensionality reduction processing on the multi-dimensional data to obtain the dimensionality reduction data of the multi-dimensional data.

[0122] Multiple hidden layers for extracting abstract features through linear transformation.

[0123] Two output layers, where one output layer is used to output the flood prevention prediction result using an activation function based on the abstract features, and the other output layer is used to output the flow speed prediction result of the flood prevention transfer control points based on the abstract features output by some of the hidden layers.

[0124] An adjustment unit for adjusting the bias matrix corresponding to the partial hidden layers according to the result calculated by the turbulence model.

[0125] The model establishment device for rail transit flood prevention provided by the embodiments of the present invention can execute the model establishment method for rail transit flood prevention provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0126] Embodiment 4

[0127] Figure 4 It is a schematic structural diagram of a server provided by Embodiment 4 of the present invention. Figure 4 It shows a block diagram of an exemplary server 12 suitable for implementing the embodiments of the present invention. Figure 4 The displayed server 12 is only an example and should not bring any limitation to the functions and usage scope of the embodiments of the present invention.

[0128] As Figure 4 shown, the server 12 is presented in the form of a general-purpose computing device. The components of the server 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 connecting different system components (including the system memory 28 and the processing unit 16).

[0129] Bus 18 represents one or more of several types of bus architectures, including a memory bus or memory controller, a peripheral bus, an Accelerated Graphics Port, a processor bus, or a local bus using any of a variety of bus architectures. By way of example, and not limitation, these architectures include Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus.

[0130] Server 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by server 12, including both volatile and nonvolatile media, removable and non-removable media.

[0131] System memory 28 can include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. Server 12 may further include other removable / non-removable, volatile / nonvolatile computer system storage media. By way of example only, storage system 34 can be used for reading and writing non-removable, nonvolatile magnetic media ( Figure 4 not shown and typically called a "hard disk drive"). Although Figure 4 not shown in the figures, a disk drive for reading and writing removable, nonvolatile magnetic disks (such as a "floppy disk"), and an optical disk drive for reading and writing removable, nonvolatile optical disks (such as a CD-ROM, DVD-ROM, or other optical media) can be provided. In each case, the drive can be connected to bus 18 by one or more data media interfaces. Memory 28 can include at least one program product having a set (e.g., at least one) of program modules that are configured to carry out the functions of embodiments of the present invention.

[0132] A program / utility 40 having a set (at least one) of program modules 42 can be stored in, for example, memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data, each of which examples or some combination thereof may include an implementation of a network environment. The program modules 42 typically carry out the functions and / or methods of the embodiments described herein.

[0133] Server 12 can also communicate with one or more external devices 14 (such as keyboards, pointing devices, monitors 24, etc.), and can also communicate with one or more devices that enable users to interact with the server 12, and / or communicate with any device that enables the server 12 to communicate with one or more other computing devices (such as network cards, modems, etc.). Such communication can be carried out through the input / output (I / O) interface 22. In addition, the server 12 can also communicate with one or more networks (such as local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) through the network adapter 20. As shown in the figure, the network adapter 20 communicates with other modules of the server 12 through the bus 18. It should be understood that although not shown in the figure, other hardware and / or software modules can be used in combination with the server 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0134] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the model establishment method for flood prevention in rail transit provided by the embodiments of the present invention.

[0135] Embodiment 5

[0136] Embodiment 5 of the present invention also provides a storage medium containing computer-executable instructions, and the computer-executable instructions are used to execute any one of the model establishment methods for flood prevention in rail transit provided by the above embodiments when executed by a computer processor.

[0137] The computer storage medium of the embodiments of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (non-exhaustive list) of the computer-readable storage medium include: an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, the computer-readable storage medium can be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, apparatus, or device.

[0138] A computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device.

[0139] The program code contained on a computer-readable medium may be transmitted using any appropriate medium, including - but not limited to - wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0140] The computer program code for performing the operations of the present invention may be written in one or more programming languages or combinations thereof. The programming languages include object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or device. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect through the Internet).

[0141] Note that the above is only the preferred embodiment of the present invention and the technical principles applied. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein. Various obvious changes, re-adjustments, and substitutions can be made by those skilled in the art without departing from the protection scope of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments. Without departing from the concept of the present invention, more other equivalent embodiments may be included, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A model building method for rail transit flood prevention, characterized in that: include: Acquire digital twin data of rail transit stations, and determine the number of flood prevention key points and corresponding flood prevention attribute categories based on the digital twin data, wherein the flood prevention attribute categories include flood prevention input control key points, flood prevention transitive key points, flood prevention decisive key points, and flood prevention control key points; When the total number of flood prevention input control key points, flood prevention transitive key points, flood prevention decisive key points and flood prevention control key points is greater than a first threshold, a dimensionality reduction improved feedforward neural network is established, the dimensionality reduction improved feedforward neural network takes the data of the flood prevention input control key points and the flood prevention control key points as input, and outputs the prediction data of the flood prevention transitive key points and the flood prevention decisive key points; When the sum of one quarter of the number of the flood prevention input control key points, the number of flood prevention transitive key points, the number of flood prevention decisive key points and the number of flood prevention control key points is less than the first threshold, a GNN model is established, and the GNN model establishes and updates the connection relationship between hierarchical nodes according to the hierarchical relationship of the clustering nodes, flood prevention transitive key points, flood prevention decisive key points and flood prevention control key points of the flood prevention input control key points, and continuously updates each hierarchical node according to the data of the flood prevention input control key points, and obtains the predicted classification of the flood prevention transitive key points and the flood prevention decisive key points according to the updated hierarchical nodes.

2. The model building method for rail transit flood prevention according to claim 1, characterized in that: The GNN model includes: An aggregation layer, used to establish an input node graph according to the positional relationship of the flood prevention input control key points, and generate flood prevention input aggregation nodes using an aggregation function; A graph building layer, for generating a directed graph using the positions of the flood control input aggregation nodes, flood control transitive key points, flood control decisive key points and flood control key points and the water flow sequence relationship, wherein the flood control key points are negative; The fully connected layer is used to utilize the attributes of the flood prevention transitive key points and the flood prevention decisive key points output by the graph building layer to output the flood prevention possibility probability of the flood prevention transitive key points and the flood prevention decisive key points.

3. The model building method for rail transit flood prevention according to claim 1, characterized in that: The obtaining of digital twin data of rail transit stations includes: Obtain virtual precipitation data predicted by large meteorological models; Determine the corresponding flood prevention management strategy according to the virtual precipitation data, and update the digital twin data based on the flood prevention management strategy.

4. The model building method for rail transit flood prevention according to claim 3 is characterized in that: Determining the corresponding flood prevention management strategy according to the virtual precipitation data, and updating the digital twin data based on the flood prevention management strategy, includes: Adjust the height of water retaining facilities and the drainage power of peripheral drainage facilities according to precipitation data and the comparison table of water retaining facilities and peripheral drainage facilities; Update the digital twin data based on the adjusted water retaining facilities and peripheral drainage facilities.

5. The model building method for rail transit flood prevention according to claim 1, characterized in that: Determining the number of flood prevention key points and corresponding flood prevention attribute categories based on digital twin data includes: Determine each flood prevention attribute category according to flood prevention requirements, and select several flood prevention points according to flood prevention transfer key points, flood prevention decisive key points and flood prevention control key points based on digital twin data; Determine the redundancy index and hot standby index of each flood prevention point based on a number of flood prevention points; Calculating the vulnerability of each flood prevention point according to the redundancy index and the hot standby index, and calculating the stability of each flood prevention point; According to the fragility and stability, key flood prevention points are selected from the flood prevention points.

6. The model building method for rail transit flood prevention according to claim 5, characterized in that: Several flood prevention points are selected according to each category based on the digital twin data, including: Divide the stairs within the site into grid units based on the site digital twin data; Grid cells were selected from the bottom of the stairs, the side walls of the middle floor, and the middle floor positions as the key points of flood prevention transferability; The water velocity at each flood control transfer key point is calculated using the turbulence model based on the data of the flood control input control key points.

7. The model building method for rail transit flood prevention according to claim 6, characterized in that: The dimension reduction improved feedforward neural network comprises: The input layer is used to convert the flood control input control key points and the locations of the flood control key points, the predicted precipitation, the time of the predicted precipitation, the spatial location data, the redundancy index and the hot standby index into multi-dimensional data of the flood control points; The dimension reduction layer is used to perform dimension reduction processing on the multi-dimensional data of the flood prevention points to obtain dimension reduction data of the multi-dimensional data of the flood prevention points; Multiple hidden layers for extracting abstract features through linear transformation; Two output layers, one of which is used to output flood prevention prediction results using an activation function based on abstract features, and the other is used to output the flow velocity prediction results of flood prevention transfer control points based on the abstract features output by some hidden layers; The adjustment unit is used to adjust the bias matrix corresponding to the part of hidden layers according to the result calculated by the turbulence model.

8. A model building device for rail transit flood prevention, characterized in that: include: An acquisition module is used to acquire digital twin data of rail transit stations, and determine the number of flood prevention key points and corresponding flood prevention attribute categories according to the digital twin data, wherein the flood prevention attribute categories include flood prevention input control key points, flood prevention transitive key points, flood prevention decisive key points and flood prevention control key points; The first establishment module is used to establish a dimensionality reduction improved feedforward neural network when the total number of flood prevention input control key points, flood prevention transitive key points, flood prevention decisive key points and flood prevention control key points is greater than a first threshold value. The dimensionality reduction improved feedforward neural network takes the data of the flood prevention input control key points and the flood prevention control key points as input, and outputs the prediction data of the flood prevention transitive key points and the flood prevention decisive key points; The second establishing module is used to establish a GNN model when the sum of the number of one-fourth of the flood prevention input control key points, the number of flood prevention transitive key points, the number of flood prevention decisive key points and the number of flood prevention control key points is less than a first threshold value. The GNN model establishes and updates the connection relationship between hierarchical nodes according to the hierarchical relationship of the clustering nodes, flood prevention transitive key points, flood prevention decisive key points and flood prevention control key points of the flood prevention input control key points, and continuously updates each hierarchical node according to the data of the flood prevention input control key points, and obtains the predicted classification of the flood prevention transitive key points and the flood prevention decisive key points according to the updated hierarchical nodes.

9. A server, characterized in that: The server comprises: one or more processors; A storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the model building method for rail transit flooding prevention as described in any one of claims 1-7.

10. A storage medium containing computer executable instructions, characterized in that: The computer executable instructions, when executed by a computer processor, are used to execute the model building method for rail transit flooding prevention as described in any one of claims 1-7.

Images