Method and apparatus for constructing an event prediction model of a multi-infrastructure system

Abstract

This invention discloses a method and apparatus for constructing an event prediction model for a multi-infrastructure system. The method involves collecting basic data from each infrastructure component within the system; simulating the basic data of each infrastructure component to obtain operational data; coupling the spatiotemporal graph networks constructed from the infrastructure components to establish a spatiotemporal graph network model; and performing transfer training on the spatiotemporal graph network model based on the operational data and real-time monitoring data to obtain the event prediction model. The event prediction model constructed using this method can scientifically and accurately assess and predict the response and evolution of the multi-infrastructure system during events, and assist in the response and handling of urban disaster events.

Description

Technical Field

[0001] This invention relates to the field of smart city management technology, and in particular to a method and apparatus for constructing an event prediction model for a multi-infrastructure system. Background Technology

[0002] Infrastructure is the core pillar for maintaining the normal operation and function of a city, and its performance is directly related to the city's safety and resilience.

[0003] Under the impact of various events, the coupling relationship between different infrastructure systems (such as stormwater drainage networks, transportation networks, and power supply networks) has a crucial influence on the propagation path and scope of disasters. This coupling relationship significantly affects the overall performance of infrastructure, especially when facing disasters and other events, where the interaction between different systems may lead to a cascading effect that exacerbates the impact of such events. However, existing simulation models often focus only on a single infrastructure system, neglecting the complex spatiotemporal coupling relationships between multiple infrastructure networks. This makes it difficult to scientifically and accurately assess and predict the response and evolution of multiple infrastructure systems during events, hindering the response and handling of urban disaster events. Summary of the Invention

[0004] To address the aforementioned shortcomings, this invention provides a method and apparatus for constructing event prediction models for multi-infrastructure systems. The provided event prediction models can scientifically and accurately assess and predict the response and evolution of multi-infrastructure systems during events, thereby assisting in the response and handling of urban disaster events.

[0005] This invention provides a method for constructing an event prediction model for a multi-infrastructure system, the method comprising:

[0006] Collect basic data from various infrastructures within a multi-infrastructure system;

[0007] Simulations were performed on the basic data of each infrastructure component to obtain operational data for each component.

[0008] Couple the spatiotemporal graph networks constructed from various infrastructure components to establish a spatiotemporal graph network model;

[0009] The spatiotemporal graph network model is transferred and trained based on the operational data and the real-time data obtained from monitoring to obtain an event prediction model.

[0010] Preferably, basic data of each infrastructure in a multi-infrastructure system is collected, including:

[0011] Collect infrastructure network data, urban land use data, population statistics, building data, meteorological data, and flood disaster data within the area to be predicted as basic data for urban facilities;

[0012] Collect pipeline attributes, manhole information, urban river information, and outlet location information as basic data for stormwater pipe networks; collect power grid structure information, electricity price information, population energy demand pattern information, building energy efficiency data, and distributed energy system information as basic data for power supply networks;

[0013] Collect road network information, vehicle information, public transportation information, and traffic checkpoint information as the basic data for the traffic network;

[0014] The collected basic data on urban facilities, stormwater pipe networks, power supply networks, and transportation networks are standardized to obtain the basic data for each infrastructure.

[0015] Preferably, the basic data of each infrastructure are simulated separately to obtain the operational data of each infrastructure, including:

[0016] The input data is determined based on the basic data of the stormwater pipe network. The stormwater pipe network and the urban river network are simulated using a preset storm runoff model to obtain the corresponding output data.

[0017] The input data is determined based on the basic data of the traffic network, and the traffic network is simulated using a preset multi-agent traffic simulation model to obtain the corresponding output data.

[0018] The input data is determined based on the basic data of the power supply network, and the power supply network is simulated using a preset urban energy analysis model to obtain the corresponding output data.

[0019] The obtained output data and the corresponding input data are used as the operating data of the rainwater pipe network.

[0020] Furthermore, the input data of the storm runoff model includes rainfall data, topographic data, land use data, pipeline structure data, river data, and initial condition data, and the corresponding output data includes pipeline flow rate, pipeline water level, and surface water depth.

[0021] The input data of the multi-agent traffic simulation model includes road network data, population travel data, vehicle information, public transportation data, and road water accumulation information, and the corresponding output data includes traffic flow and speed distribution.

[0022] The input data for the urban energy analysis model includes building energy consumption data, power supply network data, energy prices, population energy demand pattern information, distributed energy system information, and road waterlogging information. The corresponding output data includes power supply capacity and energy consumption.

[0023] Preferably, the spatiotemporal graph networks of each infrastructure are coupled to establish a spatiotemporal graph network model, including: modeling the graph structure according to the spatial location of the rainwater pipe network, transportation network and power supply network respectively, and establishing a surface elevation network coupled with the facilities based on the surface elevation data;

[0024] Construct a data embedding layer that encodes the data in the graph network and the surface elevation network;

[0025] A temporal learning layer is constructed based on a preset temporal model to capture the temporal dependencies of the data embedding layer, and a spatial learning layer is constructed based on a preset self-attention mechanism to capture the spatial topological associations of the data embedding layer.

[0026] A fully connected layer is constructed to map the spatiotemporal features output by the temporal learning layer and the spatial learning layer into prediction metrics, thereby completing the construction of the spatiotemporal graph network model.

[0027] Furthermore, graph structures are modeled based on the spatial locations of the stormwater drainage network, transportation network, and power supply network, respectively, and a coupled surface elevation network for these facilities is established based on surface elevation data, including:

[0028] Using the spatial locations of the manholes, overflow outlets, outlets, pipe junctions, and river junctions of the stormwater pipe network as nodes, the stormwater pipes and river connection nodes as edges in the graph, and the parameters of each facility in the stormwater pipe network as attributes of the nodes and edges, a graph network of the stormwater pipe network is constructed.

[0029] A graph network for the transportation network is constructed using intersections of the transportation road network, transfer points between the transportation network and public transportation, public transportation transfer points, urban land, and entrances and exits of urban land as nodes, and the lines connecting roads, public transportation lines, and the center of urban land to entrances and exits as edges in the graph. The parameters of each facility in the transportation network are used as the attributes of the nodes and edges. A graph network for the power supply network is constructed using power facilities and urban land as nodes, power lines as edges, and the parameters of each facility in the power supply network as the attributes of the nodes and edges.

[0030] The surface elevation data is divided into catchment areas, and each catchment area is used as a node in the graph. Connections are established with other facilities in the catchment area of ​​other networks to construct the surface elevation network.

[0031] Preferably, constructing a data embedding layer to encode data in the graph network and the surface elevation network includes:

[0032] An attribute feature embedding layer is constructed to encode the attribute data in the graph network and the surface elevation network into a vector space; a time period embedding layer is constructed to encode the time features in the graph network and the surface elevation network.

[0033] A fully connected layer is used to embed time-series data segmented according to time in the graph network and the surface elevation network as a spatiotemporal adaptation embedding layer.

[0034] Preferably, the spatiotemporal graph network model is transferred and trained based on the operational data and the real-time data obtained from monitoring to obtain an event prediction model, including:

[0035] The running data is used as a simulation dataset, and the training set and validation set are divided according to a preset ratio;

[0036] Set the initial hyperparameters for transfer learning, including learning rate, batch size, network depth, and number of attention heads;

[0037] The spatiotemporal graph network model is used to learn the spatiotemporal correlation patterns of the training set, and the prediction error of the spatiotemporal graph network model is tested based on the validation set to see if it meets the preset accuracy requirements.

[0038] When the accuracy requirements are not met, the network structure or hyperparameters of different network layers in the spatiotemporal graph network model are adjusted, retrained, and the prediction error of the spatiotemporal graph network model is re-checked to see if it meets the preset accuracy requirements. When the accuracy requirements are met, the network structure and hyperparameters of the current network layer are saved as the initial event prediction model. Real-time data of various infrastructures are acquired by sensors deployed in the city as the real-time training set. The initial event prediction model is iterated and trained a preset number of times using a reset learning rate to obtain the event prediction model.

[0039] Further, based on the validation set, detecting whether the prediction error of the spatiotemporal graph network model meets the preset accuracy requirements includes:

[0040] Based on the input data in the validation set, the spatiotemporal graph network model is used to calculate the predicted values, and the indicators of the predicted values ​​include predicted flow rate, predicted water depth, predicted traffic speed, and predicted power load.

[0041] Based on the preset index measurement model, the difference between the predicted value of the spatiotemporal graph network model and the corresponding index of the output data in the validation set is calculated to obtain the flow error, water depth error, traffic speed error and power load error respectively.

[0042] When the flow rate error, the water depth error, the traffic speed error, or the power load error does not meet the corresponding tolerance range, it is determined that the accuracy requirement is not met.

[0043] When the flow rate error, the water depth error, the traffic speed error, and the power load error all meet the corresponding tolerance ranges, it is determined that the accuracy requirements are met.

[0044] This invention also provides an event prediction model construction device for a multi-infrastructure system. The device includes: a data acquisition module for acquiring basic data of each infrastructure in the multi-infrastructure system.

[0045] The simulation module is used to simulate the basic data of each infrastructure to obtain the operational data of each infrastructure; the coupling module is used to couple the spatiotemporal graph network constructed by each infrastructure to establish a spatiotemporal graph network model.

[0046] The training module is used to perform transfer training on the spatiotemporal graph network model based on the running data and the real-time data obtained from monitoring, so as to obtain the event prediction model.

[0047] The present invention provides a method and apparatus for constructing an event prediction model for a multi-infrastructure system. This method involves: collecting basic data from each infrastructure component within the system; simulating the basic data of each infrastructure component to obtain operational data; coupling the spatiotemporal graph networks constructed from the infrastructure components to establish a spatiotemporal graph network model; and performing transfer training on the spatiotemporal graph network model based on the operational data and real-time monitoring data to obtain an event prediction model. The event prediction model constructed by this application can scientifically and accurately assess and predict the response and evolution of a multi-infrastructure system during events, and assist in the response and handling of urban disaster events. Attached Figure Description

[0048] Figure 1 This is a flowchart illustrating a method for constructing an event prediction model for a multi-infrastructure system according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of the event prediction model construction device for a multi-infrastructure system provided in an embodiment of the present invention. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] Urban infrastructure systems are characterized by their wide spatial distribution, complex network topology, and significant cross-system interactions. For example, in stormwater drainage networks, rainfall-induced runoff affects water levels and flow rates, and indirectly impacts transportation and power systems through water accumulation; in transportation networks, water accumulation blocks roads, reduces traffic capacity, and further affects logistics and residents' lives; in power systems, water damage to equipment can lead to localized or grid-wide power outages, thereby exacerbating the functional failure of other infrastructure.

[0051] Accurately describing these network effects and revealing the multi-level coupling and co-evolution mechanisms among the networks are urgent problems that need to be solved in event prediction and risk management.

[0052] See Figure 1 This is a flowchart illustrating a method for constructing an event prediction model for a multi-infrastructure system according to an embodiment of the present invention. The method includes steps S1 to S4:

[0053] Step S1: Collect basic data of each infrastructure in the multi-infrastructure system;

[0054] Step S2: Simulate the basic data of each infrastructure to obtain the operational data of each infrastructure.

[0055] Step S3: Couple the spatiotemporal graph networks constructed by each infrastructure to establish a spatiotemporal graph network model;

[0056] Step S4: Based on the running data and the real-time data obtained from monitoring, perform transfer training on the spatiotemporal graph network model to obtain the event prediction model.

[0057] In the specific implementation of this embodiment, urban disaster management requires a widely applicable model. However, the complexity of the urban environment and the limitations of monitoring resources bring the following challenges to model development: 1) High difficulty in data acquisition: In real cities, multi-source data (such as rainwater pipe network flow, power consumption, traffic flow, etc.) often have problems such as missing, heterogeneous, and inconsistent data. It is difficult to comprehensively describe the system behavior by directly relying on monitoring data; 2) Huge data scale: City-level data has high spatiotemporal resolution and complex network structure. The cost of training models directly based on monitoring data is high and the computational load is huge; 3) Insufficient model transfer adaptability: Existing models usually perform well in specific scenarios, but lack cross-regional or dynamic adaptability, making it difficult to be effectively applied to actual disaster management.

[0058] To address these challenges, there is an urgent need to develop a transfer learning method that combines simulated data with monitoring data. The first step is to collect basic data from each infrastructure component within the multi-infrastructure system.

[0059] It should be noted that infrastructure may include stormwater drainage networks, power supply networks, and transportation networks. By coupling stormwater drainage networks, power supply networks, and transportation networks, it is beneficial for cities to resist extreme flooding disasters.

[0060] Model each infrastructure based on its individual infrastructure data simulation model, and generate infrastructure operation data.

[0061] After obtaining the operational data of each infrastructure, the spatiotemporal graph network constructed by each infrastructure is coupled to establish a spatiotemporal graph network model. For the method of coupled modeling of multiple infrastructure systems (such as rainwater pipe network, transportation network and power supply network), the spatial distribution characteristics and dynamic interaction processes of multiple systems are incorporated into a unified spatiotemporal graph network framework.

[0062] The spatiotemporal graph network model aims to capture the characteristics of the spatiotemporal coupling and evolution of urban infrastructure systems. Due to the massive scale and significant spatiotemporal heterogeneity of city-level multi-network data, training a high-precision model using only real monitoring data is often insufficient in a single operation. Therefore, this paper proposes a transfer learning approach, combining simulated data with real-time monitoring data to train the spatiotemporal graph network model. This approach rapidly improves the model's adaptability and robustness to complex scenarios.

[0063] This application reveals the spatiotemporal coupling relationships between infrastructure systems by dynamically modeling the interconnected effects of disasters, thus contributing to a deeper understanding of the impact mechanisms of disasters on multiple urban systems. The constructed event prediction model can scientifically and accurately assess and predict the response and evolution of multiple infrastructure systems during events. This not only provides a scientific basis for the planning and construction of urban infrastructure but also supports disaster emergency management and disaster resilience optimization, thereby enhancing the overall resilience and resource allocation efficiency of cities.

[0064] In another embodiment of the present invention, the basic data collection in step S1 specifically includes the following steps:

[0065] First, basic data on urban facilities are collected, specifically: infrastructure network data, urban land use data (land use classification, building density, green space distribution), demographic data (population density, age structure, employment status, travel survey data, daily activity patterns), building data, meteorological data, economic data, and flood disaster data within the study area.

[0066] Collect basic data on stormwater pipe networks. For stormwater pipe network data, it is necessary to be detailed down to pipe attributes (pipe diameter, slope, elevation of starting and ending points, material and its Manning roughness coefficient), manholes (location coordinates, elevation, type), urban waterways (cross-sectional shape, slope, elevation, material and its roughness coefficient), and discharge outlet locations.

[0067] Collect basic data on the power supply network. For the power supply network, it is necessary to be detailed down to the grid structure (location and corresponding equipment parameters of substations, transmission lines, and distribution lines, electricity price information, population energy demand patterns (electricity load curves of residential, commercial and industrial users), building energy efficiency data (energy consumption level of buildings, power of electrical equipment), and distributed energy systems (installed capacity and distribution of solar energy).

[0068] Collect basic data on the transportation network. For the transportation network, it is necessary to be detailed down to the road network (road grade, number of lanes, width), vehicle information (vehicle type, quantity, price, fuel type and price), public transportation (bus routes, stops, schedules, operating hours), and traffic monitoring points (location and data of traffic flow monitoring equipment).

[0069] It should be noted that in this embodiment, rainwater pipe network, power supply network, and transportation network are used as infrastructure. In other embodiments, the infrastructure can be installed and adjusted according to the predicted demand, or other basic data of infrastructure can be used.

[0070] Then, the acquired multi-source data is standardized: 1) Invalid data removal: According to data specifications, missing, duplicate, or abnormal data are checked and removed. 2) Spatial geographic coordinate registration: All spatial data are unified to the same coordinate system (such as CGCS2000) to ensure spatial consistency between different datasets. 3) Attribute data standardization: The attribute field names, units, and formats of each dataset are standardized to ensure data consistency and comparability.

[0071] By constructing a spatiotemporal graph network model that integrates geographic information (such as surface elevation and catchment area) and network topology characteristics (such as node features and edge weights), and by dynamically modeling the functional loss of nodes (such as buildings, power plants, and traffic intersections) and the transmission effect of edges during disaster propagation, we can better reveal the multi-level coupling and co-evolution mechanism between networks.

[0072] In another embodiment provided by the present invention, step S2 specifically includes the following steps:

[0073] Infrastructure operation data are generated based on the data simulation models of each individual infrastructure.

[0074] When simulating stormwater pipe networks and urban river networks, the input data is determined based on the basic data of the stormwater pipe network. A storm runoff model is used to simulate the stormwater pipe network and urban river network, and the corresponding output data is obtained.

[0075] It should be noted that the storm runoff model includes simulation models such as those by Mike Floor.

[0076] When simulating traffic networks, the input data is determined based on the basic data of the traffic network. A multi-agent traffic simulation model is used to simulate the traffic network and establish the relationship between road water depth and free flow of traffic.

[0077] It should be noted that the multi-agent traffic simulation model includes simulation models such as MATSim.

[0078] When simulating the power supply network, the input data is determined based on the basic data of the power supply network. The urban energy analysis model is used to simulate the power supply network, establish the relationship between the depth of water accumulation on land or roads and the switching of power supply facilities, and obtain the corresponding output data.

[0079] It should be noted that the urban energy analysis model includes simulation tools such as City Energy Analyst.

[0080] The obtained output data and the corresponding input data are used as the operating data of the stormwater pipe network, that is, the input and output data of the storm runoff model, the input and output data of the multi-agent traffic simulation model, and the input and output data of the urban energy analysis model are used as the operating data of the stormwater pipe network.

[0081] In another embodiment of the present invention, when simulating rainwater pipe network and urban river network, the input data of the storm runoff model includes rainfall data, topographic data, land use data, pipe network structure data, river data and initial condition data, and the corresponding output data includes pipe flow rate, pipe water level and surface water depth.

[0082] The data includes: rainfall data (spatial and temporal distribution data of designed rainstorm events or historical rainfall events); topographic data (DEM, used to determine surface runoff paths); land use data (surface parameters for different land use types, such as permeability and roughness coefficient); pipeline network structure data (spatial location and attribute parameters of pipelines, manholes, overflow outlets, etc.); river data (spatial location, cross-sectional shape, and attribute parameters of rivers); initial conditions (initial water level of the pipeline network, initial soil moisture content, etc.); pipeline flow rate and water level in the output data (changes in flow rate and water level within the pipeline at each time step); and surface water depth (changes in surface water depth in a specific area over time).

[0083] The input data for the multi-agent traffic simulation model includes road network data, population travel data, vehicle information, public transportation data, and road water accumulation information. The corresponding output data includes traffic flow and speed distribution.

[0084] The road network data includes information on road nodes and links, such as the length, capacity, free-flow speed, number of lanes, and permitted traffic modes for each link. Population travel data includes activity plans for each agent, including activity type, start and end times and locations, and mode of transport selection. Vehicle information includes vehicle type, quantity, price, fuel type and price, etc. Public transportation data includes bus routes, stops, frequency, and operating hours. Road flooding information includes road flooding depth data obtained from stormwater drainage network simulations. Traffic flow in the output data includes traffic flow for each link in the road network at each time step. Speed ​​distribution includes the average vehicle speed for each link in the road network at each time step.

[0085] The input data for the urban energy analysis model includes building energy consumption data, power supply network data, energy prices, population energy demand pattern information, distributed energy system information, and road waterlogging information. The corresponding output data includes power supply capacity and energy consumption.

[0086] The data includes: building energy consumption data (features such as building type, area, power consumption of electrical equipment, and energy efficiency rating); power supply network data (location, capacity, and connections of substations, transmission lines, and distribution lines); energy price information (electricity prices at different times); population energy demand pattern information (electricity load curves for residential, commercial, and industrial users); distributed energy system information (installed capacity and distribution of solar and wind power); and road flooding information (road flooding depth data obtained from stormwater drainage network simulations, used to assess the impact on power facilities).

[0087] The output data includes power supply capacity, which refers to the power supply capacity of the power supply network facilities. Energy consumption refers to the total energy consumption and distribution among various users.

[0088] The spatiotemporal dataset obtained by integrating the input data and simulation output of each model includes five types of data: meteorological data, namely rainfall data; flow and water level data of rainwater pipes and urban river networks; flow and speed data of transportation networks; power supply capacity and user consumption data of power supply networks; and surface water accumulation data.

[0089] In another embodiment provided by the present invention, step S3 specifically includes the following steps:

[0090] Modeling of spatiotemporal graph networks is performed based on various infrastructures, including the design of graph structures and spatiotemporal graph network structures.

[0091] When designing the graph structure, the graph structure is modeled according to the spatial location of the rainwater pipe network, transportation network and power supply network respectively, and a surface elevation network coupled with the facilities is established based on the surface elevation data.

[0092] After the coupled infrastructure network graph is constructed, the spatiotemporal graph structure is designed, and a suitable graph neural network structure is designed to learn the dynamic spatiotemporal processes that occur on the graph.

[0093] A data embedding layer is constructed to encode the data in the graph network and the surface elevation network.

[0094] A temporal learning layer is constructed, which is divided into a temporal learning layer and a spatial learning layer. The temporal learning layer is constructed based on a Transformer or LSTM temporal model to capture temporal dependencies. The spatial learning layer is constructed by using a GCN (Graph Convolutional Network) or a Transformer-based self-attention mechanism to capture spatial topological relationships.

[0095] A fully connected layer is constructed to map the spatiotemporal features output by the temporal learning layer and the spatial learning layer into prediction metrics, which are then used as the output of the prediction layer, thus completing the construction of the spatiotemporal graph network model.

[0096] This application innovatively proposes a method for coupled modeling of multiple infrastructure systems (such as stormwater drainage networks, transportation networks, and power supply networks), incorporating the spatial distribution characteristics and dynamic interaction processes of multiple systems into a unified spatiotemporal graph network framework. This method can integrate the node and edge characteristics of each system, accurately characterize the complex relationships between systems, and provide an effective tool for multi-network collaborative analysis.

[0097] In another embodiment of the present invention, when modeling the graph structure and establishing a facility-coupled surface elevation network based on surface elevation data, the specific steps include:

[0098] For the stormwater pipe and urban waterway network, the spatial locations of stormwater system manholes, overflow outlets, outlets, pipe junctions, and waterway junctions are abstracted as nodes. The connections between stormwater pipes and waterways are used as edges in the graph, and the parameters of each facility are used as attributes of nodes and edges to construct the stormwater pipe network.

[0099] For the transportation network, nodes are abstracted from intersections of transportation roads, transfer points between the transportation network and public transportation, public transportation transfer points, urban land, and entrances / exits of urban land. The graph network of the transportation network is constructed using the lines connecting roads, public transportation routes, and the center of urban land to entrances / exits as edges, and the parameters of each facility as attributes of nodes and edges.

[0100] For a power supply network, power facilities and urban land use are abstracted as nodes. Power lines are treated as edges in the graph, and the parameters of each facility are used as attributes of nodes and edges to construct the graph network of the power supply network.

[0101] For a surface elevation network, the surface elevation data is divided into catchment areas. Each catchment area is used as a point on the map, and connections are established with other facilities (points) in the spatial location of the catchment area of ​​other networks. This connection is the coupling relationship between different infrastructure networks, thus constructing the surface elevation network.

[0102] In another embodiment of the present invention, the constructed data embedding layer includes an attribute feature embedding layer, a periodicity (daily, weekly, yearly) embedding layer, and a spatio-temporal adaptive embedding layer.

[0103] The attribute data of nodes in the graph network and the surface elevation network are encoded into a vector space through the attribute feature embedding layer.

[0104] By constructing a time-period embedding layer, the time features of the graph network and the surface elevation network, such as daily, weekly, and annual periods, are encoded.

[0105] By constructing a spatiotemporally adaptive embedding layer: a fully connected layer (FC(·)) is used to embed the simulated time-series data segmented according to the time period.

[0106] This invention reveals the spatiotemporal coupling relationships between infrastructure systems by dynamically modeling the interconnected effects of disasters, thus helping to deepen the understanding of the impact mechanisms of disasters on multiple urban systems. This capability can not only provide a scientific basis for the planning and construction of urban infrastructure, but also support disaster emergency management and disaster resilience optimization, thereby improving the overall resilience and resource allocation efficiency of cities.

[0107] In another embodiment provided by the present invention, step S4 specifically includes the following steps:

[0108] Spatiotemporal graph network models aim to capture the characteristics of the coupling and evolution of urban infrastructure systems in the spatiotemporal dimensions. Due to the massive scale and significant spatiotemporal heterogeneity of city-level multi-network data, it is often difficult to train a high-precision model in one go using only real monitoring data. Therefore, introducing the concept of transfer learning from simulated data to monitoring data can rapidly improve the model's adaptability and robustness to complex scenarios.

[0109] The data preparation section includes the simulation dataset: simulation results from previously generated stormwater drainage networks, traffic networks, and power supply networks. The monitoring dataset consists of real-time data acquired by various sensors and real-time monitoring systems (flow meters, traffic detectors, electricity meters, etc.) at the corresponding nodes.

[0110] Using simulated data, we can initially learn the spatiotemporal correlation patterns of urban stormwater drainage networks, transportation networks, and power supply networks; we can learn basic spatiotemporal characteristics (such as coupling relationships between nodes and time series patterns) to lay a solid foundation for subsequent transfer training.

[0111] The training process includes:

[0112] Data partitioning: The simulated data is divided into a training set and a validation set according to a preset ratio over time, with a portion reserved as a test set.

[0113] It should be noted that the preset ratio can be set or adjusted according to the actual situation, and can be set to 7:2:1.

[0114] Set the initial hyperparameters for transfer learning, including learning rate, batch size, network depth, and number of attention heads (if using a Transformer-based spatiotemporal learning module).

[0115] The spatiotemporal graph network model is used to learn the spatiotemporal correlation patterns of the training set, and the prediction error of the spatiotemporal graph network model is tested based on the validation set to see if it meets the preset accuracy requirements.

[0116] If the accuracy requirement is not met, the network structure or hyperparameters of different network layers in the spatiotemporal graph network model are adjusted, retrained, and the prediction error of the spatiotemporal graph network model is re-checked to see if it meets the preset accuracy requirement. If the accuracy requirement is met, the network structure and hyperparameters of the current network layer are saved as the initial event prediction model, that is, an initial event prediction model with good prediction ability in the simulated data domain is obtained.

[0117] By acquiring real-time data from various infrastructures through sensors deployed in the city, and using this data as a real-time training set, the initial event prediction model is iterated and trained a preset number of times using a reset learning rate to obtain the event prediction model.

[0118] The model is fine-tuned in stages. First, the first few layers of the spatiotemporal graph network (such as the embedding layer and some spatiotemporal learning layers) are "frozen," and only the higher layers (such as the later layers or prediction layers) are trained to avoid completely resetting the model parameters. Then, the model structure and parameters pre-trained on simulated data are retained, and the monitored data is used as a new training set. Multiple rounds of iteration are performed with a small learning rate, allowing the model to retain existing knowledge while correcting itself for real-world scenarios.

[0119] It should be noted that this case can also be used to verify and test the time prediction model. If the accuracy of the validation set continues to improve, it means that the model is gradually adapting to the real scenario. If overtraining causes the model to perform well on the monitoring data but fail on the simulated data or another part of the monitoring data, it means that the scale of the monitoring data needs to be increased.

[0120] This invention significantly reduces reliance on large-scale monitoring data by combining simulated and monitoring data and employing transfer learning techniques. By initially training the model on simulated data and then fine-tuning it using limited monitoring data, the model not only achieves high prediction accuracy but also substantially reduces the economic and time costs of data collection and processing.

[0121] In another embodiment of the present invention, the accuracy verification process specifically includes the following steps:

[0122] Based on the input data in the validation set, the spatiotemporal graph network model is used to calculate the predicted values, and the indicators of the predicted values ​​include predicted flow rate, predicted water depth, predicted traffic speed, and predicted power load.

[0123] The difference between the predicted value of the spatiotemporal graph network model and the corresponding index of the output data in the validation set is calculated based on the preset index measurement model.

[0124] The indicator measurement model includes one of the following:

[0125]

[0126]

[0127] in, It is the predicted value, y i This represents the true value, and N is the number of samples.

[0128] The root mean square error (RMSE) is sensitive to outliers and is mainly used to measure the overall level of prediction error.

[0129] Mean Absolute Error (MAE) is used to average absolute errors, reflecting the average deviation, and is often used to measure the absolute degree of prediction deviation.

[0130] Average Prediction (MAP) reflects the proportion of prediction error on the scale of the true value by relativizing the error.

[0131] By monitoring the model's performance in real time using a validation set during training (e.g., traffic flow prediction error, water depth prediction error, traffic speed prediction error, power load prediction error, etc.), and adjusting the network structure or hyperparameters if the performance does not meet expectations, the model's accuracy can be improved.

[0132] When the flow rate error, the water depth error, the traffic speed error, or the power load error does not meet the corresponding tolerance range, it is determined that the accuracy requirement is not met.

[0133] When the flow rate error, the water depth error, the traffic speed error, and the power load error all meet the corresponding tolerance ranges, it is determined that the accuracy requirements are met.

[0134] It should be noted that the indicator measurement model pointed out in this embodiment is only a preferred embodiment. In other embodiments, the indicator error can be calculated by other measurement models.

[0135] After transfer training, the model can better adapt to dynamic changes in the real world. At this point, the model can be deployed to an urban operation management platform or the cloud, and connected to real-time or periodically updated monitoring data. This embodiment of the invention also provides an event prediction model building device for multi-infrastructure systems, see [link to related documentation]. Figure 2 This is a schematic diagram of the structure of an event prediction model building device for a multi-infrastructure system provided in an embodiment of the present invention. The device includes:

[0136] The data acquisition module is used to collect basic data from various infrastructures in a multi-infrastructure system.

[0137] The simulation module is used to simulate the basic data of each infrastructure to obtain the operational data of each infrastructure; the coupling module is used to couple the spatiotemporal graph network constructed by each infrastructure to establish a spatiotemporal graph network model.

[0138] The training module is used to perform transfer training on the spatiotemporal graph network model based on the running data and the real-time data obtained from monitoring, so as to obtain the event prediction model.

[0139] It should be noted that the event prediction model building device for multi-infrastructure systems provided in the embodiments of the present invention can execute the event prediction model building method for multi-infrastructure systems described in any of the above embodiments. The specific functions of the event prediction model building device for multi-infrastructure systems will not be elaborated here.

[0140] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for constructing an event prediction model for a multi-infrastructure system, characterized in that, The method includes: Collect basic data from various infrastructures within a multi-infrastructure system; Simulations were performed on the basic data of each infrastructure to obtain the operational data of each infrastructure. Couple the spatiotemporal graph networks constructed from various infrastructure components to establish a spatiotemporal graph network model; Based on the operational data and the real-time data obtained from monitoring, the spatiotemporal graph network model is transferred and trained to obtain an event prediction model. Collect basic data from each infrastructure component in a multi-infrastructure system, including: Collect infrastructure network data, urban land use data, population statistics, building data, meteorological data, and flood disaster data within the area to be predicted as basic data for urban facilities; Collect pipeline attributes, manhole information, urban river information, and discharge outlet location information as basic data for the stormwater pipe network; Collect information on power grid structure, electricity prices, population energy demand patterns, building energy efficiency data, and distributed energy systems as basic data for the power supply network; Collect road network information, vehicle information, public transportation information, and traffic checkpoint information as the basic data for the traffic network; The collected basic data of urban facilities, stormwater pipe networks, power supply networks and transportation networks are standardized to obtain the basic data of each infrastructure. The spatiotemporal graph networks constructed from various infrastructure components are coupled to establish a spatiotemporal graph network model, including: The graph structure is modeled based on the spatial location of the rainwater pipe network, transportation network and power supply network respectively, and a surface elevation network coupled with the facilities is established based on the surface elevation data; Construct a data embedding layer that encodes the data in the graph network and the surface elevation network; A temporal learning layer is constructed based on a preset temporal model to capture the temporal dependencies of the data embedding layer, and a spatial learning layer is constructed based on a preset self-attention mechanism to capture the spatial topological associations of the data embedding layer. A fully connected layer is constructed to map the spatiotemporal features output by the temporal learning layer and the spatial learning layer into prediction metrics, thereby completing the construction of the spatiotemporal graph network model. Graph structures are modeled based on the spatial locations of stormwater drainage networks, transportation networks, and power supply networks, respectively. A coupled surface elevation network is then established based on surface elevation data, including: Using the spatial locations of the manholes, overflow outlets, outlets, pipe junctions, and river junctions of the stormwater pipe network as nodes, the stormwater pipes and river connection nodes as edges in the graph, and the parameters of each facility in the stormwater pipe network as attributes of the nodes and edges, a graph network of the stormwater pipe network is constructed. A graph network of the transportation network is constructed using the intersections of the transportation road network, the transfer points between the transportation network and public transportation, the transfer points of public transportation, urban land and the entrances and exits of urban land as nodes, the lines connecting the roads, public transportation lines and the center of urban land to the entrances and exits as edges in the graph, and the parameters of each facility in the transportation network as the attributes of the nodes and edges. A graph network of the power supply network is constructed using power facilities and urban land as nodes, power lines as edges, and the parameters of each facility in the power supply network as the attributes of the nodes and edges. The surface elevation data is divided into catchment areas, and each catchment area is used as a node in the graph. Connections are established with other facilities in the catchment area of ​​other networks to construct the surface elevation network. 2.The method of claim 1, wherein, Simulations were performed on the basic data of each infrastructure component to obtain operational data for each component, including: The input data is determined based on the basic data of the stormwater pipe network. The stormwater pipe network and the urban river network are simulated using a preset storm runoff model to obtain the corresponding output data. The input data is determined based on the basic data of the traffic network, and the traffic network is simulated using a preset multi-agent traffic simulation model to obtain the corresponding output data. The input data is determined based on the basic data of the power supply network, and the power supply network is simulated using a preset urban energy analysis model to obtain the corresponding output data. The obtained output data and the corresponding input data are used as the operating data of the rainwater pipe network. 3.The method of claim 2, wherein, The input data of the storm runoff model includes rainfall data, topographic data, land use data, pipeline structure data, river data, and initial condition data. The corresponding output data includes pipeline flow rate, pipeline water level, and surface water depth. The input data of the multi-agent traffic simulation model includes road network data, population travel data, vehicle information, public transportation data, and road water accumulation information, and the corresponding output data includes traffic flow and speed distribution. The input data for the urban energy analysis model includes building energy consumption data, power supply network data, energy prices, population energy demand pattern information, distributed energy system information, and road waterlogging information. The corresponding output data includes power supply capacity and energy consumption.

4. The method for constructing an event prediction model for a multi-infrastructure system as described in claim 1, characterized in that, Constructing a data embedding layer to encode data in the graph network and the surface elevation network includes: Construct an attribute feature embedding layer that encodes attribute data from the graph network and the surface elevation network into a vector space; Construct a time-period embedding layer that encodes the temporal features in the graph network and the surface elevation network; A fully connected layer is used to embed time-series data segmented according to time in the graph network and the surface elevation network as a spatiotemporal adaptation embedding layer.

5. The method for constructing an event prediction model for a multi-infrastructure system as described in claim 1, characterized in that, The spatiotemporal graph network model is transferred and trained based on the operational data and the real-time data obtained from monitoring to obtain an event prediction model, including: The running data is used as a simulation dataset, and the training set and validation set are divided according to a preset ratio; Set the initial hyperparameters for transfer learning, including learning rate, batch size, network depth, and number of attention heads; The spatiotemporal graph network model is used to learn the spatiotemporal correlation patterns of the training set, and the prediction error of the spatiotemporal graph network model is tested based on the validation set to see if it meets the preset accuracy requirements. When the accuracy requirements are not met, the network structure or hyperparameters of different network layers in the spatiotemporal graph network model are adjusted, retrained, and the prediction error of the spatiotemporal graph network model is re-checked to see if it meets the preset accuracy requirements. When the accuracy requirement is met, the network structure and hyperparameters of the current network layer are saved as the initial event prediction model. By acquiring real-time data from various infrastructures through sensors deployed in the city, and using this data as a real-time training set, the initial event prediction model is iterated and trained a preset number of times using a reset learning rate to obtain the event prediction model.

6. The method for constructing an event prediction model for a multi-infrastructure system as described in claim 5, characterized in that, The detection of whether the prediction error of the spatiotemporal graph network model meets the preset accuracy requirements based on the validation set includes: Based on the input data in the validation set, the spatiotemporal graph network model is used to calculate the predicted values, and the indicators of the predicted values ​​include predicted flow rate, predicted water depth, predicted traffic speed, and predicted power load. Based on the preset index measurement model, the difference between the predicted value of the spatiotemporal graph network model and the corresponding index of the output data in the validation set is calculated to obtain the flow error, water depth error, traffic speed error and power load error respectively. When the flow rate error, the water depth error, the traffic speed error, or the power load error does not meet the corresponding tolerance range, it is determined that the accuracy requirement is not met. When the flow rate error, the water depth error, the traffic speed error, and the power load error all meet the corresponding tolerance ranges, it is determined that the accuracy requirements are met.

7. A device for constructing an event prediction model for a multi-infrastructure system, characterized in that, The device includes: The data acquisition module is used to collect basic data from various infrastructures in a multi-infrastructure system. The simulation module is used to simulate the basic data of each infrastructure to obtain the operational data of each infrastructure. The coupling module is used to couple the spatiotemporal graph networks constructed from various infrastructures and establish a spatiotemporal graph network model. The training module is used to perform transfer training on the spatiotemporal graph network model based on the running data and the real-time data obtained by monitoring, so as to obtain the event prediction model. in, Collect basic data from each infrastructure component in a multi-infrastructure system, including: Collect infrastructure network data, urban land use data, population statistics, building data, meteorological data, and flood disaster data within the area to be predicted as basic data for urban facilities; Collect pipeline attributes, manhole information, urban river information, and discharge outlet location information as basic data for the stormwater pipe network; Collect information on power grid structure, electricity prices, population energy demand patterns, building energy efficiency data, and distributed energy systems as basic data for the power supply network; Collect road network information, vehicle information, public transportation information, and traffic checkpoint information as the basic data for the traffic network; The collected basic data of urban facilities, stormwater pipe networks, power supply networks and transportation networks are standardized to obtain the basic data of each infrastructure. The spatiotemporal graph networks constructed from various infrastructure components are coupled to establish a spatiotemporal graph network model, including: The graph structure is modeled based on the spatial location of the rainwater pipe network, transportation network and power supply network respectively, and a surface elevation network coupled with the facilities is established based on the surface elevation data; Construct a data embedding layer that encodes the data in the graph network and the surface elevation network; A temporal learning layer is constructed based on a preset temporal model to capture the temporal dependencies of the data embedding layer, and a spatial learning layer is constructed based on a preset self-attention mechanism to capture the spatial topological associations of the data embedding layer. A fully connected layer is constructed to map the spatiotemporal features output by the temporal learning layer and the spatial learning layer into prediction metrics, thereby completing the construction of the spatiotemporal graph network model. Graph structures are modeled based on the spatial locations of stormwater drainage networks, transportation networks, and power supply networks, respectively. A coupled surface elevation network is then established based on surface elevation data, including: Using the spatial locations of the manholes, overflow outlets, outlets, pipe junctions, and river junctions of the stormwater pipe network as nodes, the stormwater pipes and river connection nodes as edges in the graph, and the parameters of each facility in the stormwater pipe network as attributes of the nodes and edges, a graph network of the stormwater pipe network is constructed. A graph network of the transportation network is constructed using the intersections of the transportation road network, the transfer points between the transportation network and public transportation, the transfer points of public transportation, urban land and the entrances and exits of urban land as nodes, the lines connecting the roads, public transportation lines and the center of urban land to the entrances and exits as edges in the graph, and the parameters of each facility in the transportation network as the attributes of the nodes and edges. A graph network of the power supply network is constructed using power facilities and urban land as nodes, power lines as edges, and the parameters of each facility in the power supply network as the attributes of the nodes and edges. The surface elevation data is divided into catchment areas, and each catchment area is used as a node in the graph. Connections are established with other facilities in the catchment area of ​​other networks to construct the surface elevation network.

Images