Sewage pipe network visualization method, device, equipment and medium

Abstract

The invention relates to a sewage pipe network visualization method, device and equipment and a medium. The method comprises the following steps: acquiring a topological structure, physical space parameters and real-time monitoring data of a sewage pipe network; constructing a pipe network simulation model based on the topological structure and the physical space parameters; based on the real-time monitoring data and the pipe network simulation model, determining an abnormal identification threshold value of each pipe section in the pipe network simulation model to obtain a pipe network monitoring model; based on the pipe network monitoring model, in combination with real-time monitoring data, pipe network abnormity identification is carried out, and in response to the identified pipe network abnormity, a potential abnormal pipe section is positioned; based on the pipe network monitoring model, a pipe network scheduling scheme is generated in combination with the abnormal pipe sections; and integrating the real-time monitoring data, the pipe network monitoring model, the potential abnormal pipe section and the pipe network scheduling scheme to generate a visual result. By adopting the method, the accuracy and efficiency of sewage pipe network anomaly identification can be improved, and rapid positioning and intelligent scheduling are realized.

Description

Technical Field

[0001] This invention belongs to the field of smart water management, and in particular relates to a method, device, equipment and medium for visualizing sewage pipe networks. Background Technology

[0002] With the development of smart water management technology, intelligent monitoring technology for sewage pipe networks has emerged. This technology integrates sensor networks and data analysis platforms to achieve real-time monitoring and data collection of the operational status of sewage pipe networks, featuring automation and high real-time performance. Current sewage pipe network monitoring methods, or traditional approaches, typically handle pipe network anomalies by combining manual inspections with fixed threshold alarms. This relies on periodic on-site checks or comparisons of basic sensor data, focusing on local analysis of single pipe segments and lacking global topological correlation. Current traditional methods suffer from low accuracy in anomaly identification, low efficiency, difficulty in quickly locating abnormal pipe segments, and a lack of intelligent scheduling solutions, leading to response delays and resource waste. Summary of the Invention

[0003] Therefore, it is necessary to provide a method, device, equipment, and medium for visualizing sewage pipe networks that can solve the above problems.

[0004] Firstly, this application provides a method for visualizing sewage pipe networks, including:

[0005] Obtain the topology, physical spatial parameters, and real-time monitoring data of the sewage pipe network;

[0006] Based on topology and physical space parameters, a graph neural network is used to construct a pipeline network simulation model;

[0007] Based on real-time monitoring data and pipeline network simulation model, the anomaly identification threshold of each pipe segment in the pipeline network simulation model is determined by particle swarm optimization algorithm, thus obtaining the pipeline network monitoring model;

[0008] Based on the pipeline network monitoring model and combined with real-time monitoring data, pipeline network anomalies are identified, and in response to the identification of pipeline network anomalies, potential abnormal pipe sections are located through backtracking algorithms.

[0009] Based on the pipeline network monitoring model and combined with abnormal pipeline sections, a pipeline network scheduling scheme is generated through dynamic programming algorithm.

[0010] By integrating real-time monitoring data, pipeline monitoring models, potential abnormal pipeline sections, and pipeline scheduling schemes, visualization results are generated through visualization methods.

[0011] In one embodiment, a pipeline network simulation model is constructed using a graph neural network based on topology and physical space parameters, including:

[0012] Based on the topology, a graph structure data of the sewage pipe network is constructed, in which nodes represent pipe segments and edges represent the connection relationships between pipe segments.

[0013] Based on physical space parameters, feature vectors are assigned to each node of the graph structure data to form graph structure data with node features;

[0014] Graph structure data with node features is input into a graph neural network, and node-encoded features are generated through feature encoding.

[0015] Extract the adjacency matrix from the graph structure data;

[0016] Based on node encoding features and combined with the adjacency matrix, multi-layer graph convolution propagation is used to generate propagated node features;

[0017] Based on the characteristics of the nodes after propagation, simulated hydraulic data for each pipe segment are generated using fluid dynamics methods.

[0018] Based on simulated hydraulic data, the parameters of the graph neural network are adjusted using the gradient descent algorithm to obtain a pipeline network simulation model.

[0019] In one embodiment, based on real-time monitoring data and a pipeline network simulation model, an anomaly identification threshold for each pipe segment in the pipeline network simulation model is determined using a particle swarm optimization algorithm to obtain a pipeline network monitoring model, including:

[0020] Based on real-time monitoring data and pipeline network simulation models, residual time-series data for each pipeline segment are generated through difference calculation;

[0021] Based on residual time series data and combined with preset optimization objectives, multiple candidate threshold schemes are generated by particle swarm optimization algorithm. Each candidate threshold scheme contains a set of anomaly identification thresholds corresponding to each pipe segment.

[0022] Based on the preset optimization objective, the fitness score of each candidate threshold scheme is calculated using a fitness function;

[0023] Based on the fitness score, the optimal anomaly detection threshold scheme is selected from the candidate threshold schemes;

[0024] Based on the optimal anomaly identification threshold scheme, the anomaly identification threshold for each pipe segment in the pipeline network simulation model is configured to obtain the pipeline network monitoring model.

[0025] In one embodiment, based on a pipeline monitoring model and combined with real-time monitoring data, pipeline anomalies are identified, and in response to the identification of pipeline anomalies, potential abnormal pipeline segments are located using a backtracking algorithm, including:

[0026] Based on the anomaly identification threshold of each pipe segment in the pipeline network monitoring model, the real-time monitoring data is compared with the anomaly identification threshold of the corresponding pipe segment to identify whether there are anomalies in the pipeline network.

[0027] In response to the identification of pipeline network anomalies, abnormal features are extracted from the real-time monitoring data corresponding to the pipeline network anomalies;

[0028] A bidirectional LSTM algorithm is used to obtain the time series correlation of abnormal features;

[0029] Based on time series correlation and combined with topological structure, the initial abnormal pipe segment is located by backtracking using the Dijkstra algorithm.

[0030] Based on the initial abnormal pipe segments and combined with the topology, potential abnormal pipe segments are obtained through a greedy algorithm for further screening.

[0031] In one embodiment, based on a pipeline monitoring model and considering abnormal pipeline segments, a pipeline scheduling scheme is generated using a dynamic programming algorithm, including:

[0032] Based on potential abnormal pipe sections and simulated hydraulic data, the scheduling constraints for each pipe section are determined.

[0033] Based on constraints and topology, the scheduling phases are divided and the decision nodes for each scheduling phase are defined.

[0034] Construct a stage state space based on decision nodes and constraints;

[0035] The stage state space is input into the dynamic programming algorithm to calculate the local scheduling scheme of the decision nodes in each scheduling stage;

[0036] By combining the topology, we can backtrack the local scheduling schemes of each scheduling stage to obtain the global scheduling path;

[0037] The conflicting nodes in the global scheduling path are corrected by a neighborhood search algorithm to generate an initial scheduling scheme.

[0038] Based on real-time monitoring data, the parameters of the initial scheduling scheme are optimized to obtain the pipeline scheduling scheme.

[0039] In one embodiment, based on the initial abnormal pipe segments and combined with the topology, potential abnormal pipe segments are obtained through a greedy algorithm for extended screening. The greedy algorithm for extended screening is implemented using the following formula:

[0040]

[0041] in, Let C be the pipe segment selected in the current greedy step, and let C be the pipe segment adjacent to the current potentially abnormal pipe segment P. This refers to the set of pipe segments directly connected to pipe segment v. Let U be the topological distance between pipe segments u and v, and P represent the potential abnormal pipe segment in the current step. P is initialized to the initial abnormal pipe segment.

[0042] In one embodiment, real-time monitoring data, pipeline monitoring models, potentially abnormal pipeline segments, and pipeline scheduling schemes are integrated, and visualization results are generated through visualization methods, including:

[0043] Based on real-time monitoring data, pipeline monitoring models, and potential abnormal pipe sections, extract abnormal feature data and location information of potential abnormal pipe sections;

[0044] By integrating abnormal feature data, potential abnormal pipe segment location information, and pipeline network scheduling schemes, a visual basic dataset is constructed.

[0045] Define the visualization hierarchy of each element within the basic visualization dataset;

[0046] Based on the topology, the spatial display position of each visualization level is determined through a coordinate mapping algorithm;

[0047] Based on the spatial display location and visualization hierarchy, the elements in the basic visualization dataset are integrated into the spatial display interface through dynamic annotation to form a visualization result.

[0048] Secondly, this application also provides a sewage pipe network visualization device, comprising:

[0049] The pipeline data acquisition module is used to acquire the topology, physical spatial parameters, and real-time monitoring data of the sewage pipeline network.

[0050] The simulation model building module is used to construct a pipeline network simulation model based on topology and physical space parameters using a graph neural network.

[0051] The monitoring model generation module is used to determine the anomaly identification threshold of each pipe segment in the pipeline simulation model based on real-time monitoring data and pipeline simulation model, and obtain the pipeline monitoring model by using the particle swarm optimization algorithm.

[0052] The anomaly identification and localization module is used to identify pipeline anomalies based on the pipeline monitoring model and real-time monitoring data, and to locate potential abnormal pipeline segments in response to the identification of pipeline anomalies through a backtracking algorithm.

[0053] The scheduling scheme generation module is used to generate pipeline scheduling schemes based on the pipeline monitoring model and combined with abnormal pipeline sections, using a dynamic programming algorithm.

[0054] The visualization results generation module is used to integrate real-time monitoring data, pipeline monitoring models, potential abnormal pipeline sections, and pipeline scheduling schemes, and generate visualization results through visualization methods.

[0055] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described sewage network visualization method.

[0056] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described sewage network visualization method.

[0057] The aforementioned sewage pipe network visualization method, device, computer equipment, and storage medium provide comprehensive data support by acquiring the topology, physical spatial parameters, and real-time monitoring data of the sewage pipe network. Based on the topology and physical spatial parameters, a graph neural network is used to construct a pipe network simulation model, capturing the complex spatial relationships and fluid characteristics of the pipe network and improving the accuracy of the model simulation. Based on real-time monitoring data and the pipe network simulation model, a particle swarm optimization algorithm is used to dynamically determine the anomaly identification threshold for each pipe segment, generating a pipe network monitoring model, which improves the accuracy and efficiency of anomaly identification. When a pipe network anomaly is identified based on the pipe network monitoring model and real-time monitoring data, a backtracking algorithm is used to quickly locate the potential abnormal pipe segment, enabling timely tracking of the anomaly source. Combined with the abnormal pipe segment, a dynamic programming algorithm is used to generate a pipe network scheduling scheme, optimizing resource allocation and emergency response strategies. By integrating real-time monitoring data, the pipe network monitoring model, potential abnormal pipe segments, and the scheduling scheme, intuitive visualization results are generated through visualization methods, facilitating user monitoring and decision-making. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is a flowchart of a sewage pipe network visualization method according to the present invention;

[0060] Figure 2 This is a structural diagram of a sewage pipe network visualization device according to the present invention. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0062] In one embodiment, such as Figure 1As shown, a method for visualizing sewage pipe networks is provided. This embodiment illustrates the application of this method to a terminal, but it is understood that the method can also be applied to a server, or to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. In the implementation environment, the terminal can be a monitoring workstation or mobile inspection device deployed in a water management center, integrating data reception and visualization interface display functions; the server can be a cloud computing platform used to carry complex computing tasks such as graph neural network models and particle swarm optimization algorithms. Application scenarios include: when the urban sewage pipe network system generates real-time monitoring data, sensor terminals distributed at key nodes of the pipe network upload the data to the server; based on the received topology and physical space parameters, the server constructs and updates the pipe network simulation model through a graph neural network, and dynamically calibrates the abnormal thresholds of each pipe segment using a particle swarm optimization algorithm. If abnormal situations such as abnormal pressure or sudden changes in water quality are detected, the potential abnormal pipe section is located by backtracking algorithm, and the pipeline network scheduling scheme (such as valve control, pump station start and stop) is generated by dynamic programming algorithm. The calculation results are sent to the terminal. The terminal receives the scheduling instructions and visualization data packets, and renders the pipeline network status, abnormal location and scheduling path in real time on the spatial interface through coordinate mapping and dynamic annotation technology, forming an interactive monitoring dashboard.

[0063] In this embodiment, the method includes the following steps:

[0064] S01, acquire the topology, physical spatial parameters and real-time monitoring data of the sewage pipe network.

[0065] The topology is a network diagram describing the connection relationship between pipe segments, with nodes representing pipe segments and edges representing connection relationships. The physical space parameters include geometric and material properties such as pipe diameter, length, and slope. The real-time monitoring data is time-series data collected by sensors deployed at key nodes of the pipe network, such as flow rate, pressure, or water quality indicators. It can be extracted in real time or in batches from sensor networks, geographic information systems, or historical databases through data interfaces, providing a data foundation for the visualization of sewage pipe networks.

[0066] S02, based on topology and physical space parameters, uses graph neural networks to construct a pipeline network simulation model.

[0067] This can be achieved by transforming the topology into graph data, assigning feature vectors based on physical space parameters to nodes, and inputting these vectors into a graph neural network for feature encoding and graph convolution operations. This allows the network to learn spatial relationships and dynamic characteristics, generating a network simulation model that can simulate hydraulic behavior (such as flow and pressure distribution). In practice, node features can be propagated through a multi-layer neural network and information can be aggregated by combining adjacency relationships. Parameters can be optimized based on the differences between simulated and measured data to train the model.

[0068] S03. Based on real-time monitoring data and pipeline network simulation model, the anomaly identification threshold of each pipe segment in the pipeline network simulation model is determined by particle swarm optimization algorithm to obtain the pipeline network monitoring model.

[0069] Among them, residual time-series data between measured data and simulated data of each pipe segment can be generated through difference calculation. Based on the residual data, multiple candidate schemes containing the corresponding thresholds of each pipe segment are generated using the particle swarm optimization algorithm. The comprehensive performance of each scheme is evaluated by the fitness function to select the optimal threshold combination. The optimal threshold is configured into the pipeline network simulation model to form a pipeline network monitoring model with anomaly identification capability, so as to realize the accurate perception of the abnormal state of each pipe segment.

[0070] S04, based on the pipeline network monitoring model and combined with real-time monitoring data, identifies pipeline network anomalies and, in response to the identified anomalies, locates potential abnormal pipeline segments through a backtracking algorithm.

[0071] Anomaly detection can be achieved by comparing real-time monitoring data with the anomaly identification threshold of the corresponding pipe segment. When an anomaly is detected, feature indicators are extracted from the anomaly data, and the temporal correlation of the anomaly features is obtained using a time series analysis model (such as a recurrent neural network). Combined with the pipe network topology, the initial pipe segment that may have caused the anomaly is located back by using a graph traversal algorithm (such as shortest path search). Based on the topological adjacency relationship, the set of potential anomaly pipe segments is expanded and determined by an optimization screening algorithm (such as local search), so as to achieve rapid tracking and accurate location of the source of the anomaly.

[0072] S05, based on the pipeline network monitoring model and combined with abnormal pipeline sections, generates a pipeline network scheduling scheme through a dynamic programming algorithm.

[0073] Dynamic programming is a mathematical method for multi-stage decision optimization. In its implementation, the scheduling constraints (such as flow limits or pressure ranges) of each pipe segment can be determined based on abnormal pipe segments and simulated hydraulic data. The scheduling stages are divided in combination with the pipe network topology, and the decision nodes of each stage are defined. By constructing the stage state space and applying the dynamic programming algorithm, the local scheduling scheme of each decision node is calculated. The local schemes are backtracked and integrated to obtain the global scheduling path. The conflict correction algorithm (such as neighborhood search) is used to adjust the inconsistent nodes in the path to generate the initial scheduling scheme. Based on real-time monitoring data, the parameters are optimized to form an executable pipe network scheduling scheme, realizing efficient resource allocation and emergency response for anomalies.

[0074] S06 integrates real-time monitoring data, pipeline monitoring models, potential abnormal pipe sections, and pipeline scheduling schemes, and generates visualization results through visualization methods.

[0075] This system can acquire abnormal feature data and pipe segment location information based on real-time monitoring data, pipeline monitoring models, and potentially abnormal pipe segments. It can also use data fusion methods to integrate real-time monitoring data, pipeline monitoring models, potentially abnormal pipe segments, abnormal feature data, pipe segment location information, and pipeline scheduling schemes to construct a basic visualization dataset. The display priority of each element (such as monitoring data, abnormal indications, and scheduling paths) is determined through a hierarchical partitioning algorithm. Combined with the pipeline topology, spatial mapping algorithms (such as coordinate transformation) are applied to determine the layout position of the visualization elements. Dynamic annotation technology is used to integrate the dataset into an interactive display interface, forming a visualization result that reflects the real-time status of the pipeline network, the distribution of abnormalities, and the scheduling scheme, providing intuitive support for monitoring decisions.

[0076] The aforementioned sewage pipe network visualization method constructs a comprehensive data foundation by acquiring the topology, physical spatial parameters, and real-time monitoring data of the sewage pipe network. Based on the topology and physical spatial parameters, a graph neural network is used to build a pipe network simulation model, capturing the complex spatial relationships and fluid dynamic characteristics of the pipe network, thus overcoming the limitation of traditional methods that lack global topological relationships. A particle swarm optimization algorithm combined with real-time monitoring data is used to dynamically determine the anomaly identification threshold for each pipe segment, improving the accuracy of anomaly identification and adaptability to dynamic changes in the pipe network. When an anomaly is identified, a backtracking algorithm is used to quickly locate potential abnormal pipe segments, overcoming the problems of low efficiency and slow location in traditional manual inspections. A dynamic programming algorithm is used to generate a pipe network scheduling scheme based on the abnormal pipe segments, realizing the optimization of resource allocation and intelligent emergency response. Finally, a visualization method integrates multi-source information to generate intuitive visualization results, providing real-time and comprehensive support for monitoring and decision-making.

[0077] In one embodiment, a pipeline network simulation model is constructed using a graph neural network based on topology and physical space parameters, including:

[0078] S11. Based on the topology, construct the graph structure data of the sewage pipe network, where nodes represent pipe segments and edges represent the connection relationships between pipe segments.

[0079] S12, based on physical space parameters, assigns feature vectors to each node of the graph structure data, forming graph structure data with node features;

[0080] S13, Input the graph structure data with node features into the graph neural network, and generate node encoded features through feature encoding;

[0081] S14, Extract the adjacency matrix of the graph structure data;

[0082] S15, based on node encoding features and combined with the adjacency matrix, generates propagated node features through multi-layer graph convolution propagation;

[0083] S16. Based on the characteristics of the nodes after propagation, simulated hydraulic data for each pipe section are generated using fluid dynamics methods.

[0084] S17. Based on simulated hydraulic data, the parameters of the graph neural network are adjusted using the gradient descent algorithm to obtain a pipeline network simulation model.

[0085] For example, based on the topology of the sewage pipe network, each pipe segment can be mapped to a node in the graph structure data. Simultaneously, the edges in the graph structure data correspond to the actual connection relationships between pipe segments, thus realizing the transformation from the sewage pipe network topology to the graph structure data. Physical spatial parameters of the sewage pipe network (such as pipe diameter, pipe length, slope, etc.) are extracted, quantized, and assigned to the corresponding nodes in the graph structure data. A feature vector containing the physical spatial attributes of the pipe segment is constructed for each node, forming graph structure data with node features. This graph structure data with node features is input into a pre-built graph neural network. Through feature encoding components such as embedding layers in the network, the original feature vectors of the nodes are transformed into low-dimensional dense node encoding features, achieving effective feature representation. Based on the connection relationships between nodes and edges in the graph structure data, an adjacency matrix is ​​constructed to describe the connection state between pipe segments (where a matrix element value of 1 indicates that the pipe segments represented by the corresponding two nodes are connected, and a value of 0 indicates no connection). Based on the node encoding features and combined with the extracted adjacency matrix, multiple continuous graph convolutional layers in the graph neural network (such as...) are used to... The graph convolutional network (GCN) layer performs feature propagation and aggregation. Each layer uses the adjacency matrix to obtain the features of the neighboring nodes of the current node, and then merges and calculates them with the encoded features of the current node to update the node features. After multiple layers of iterative propagation, a propagated node feature that can capture the global spatial correlation information of the pipe network is generated. Based on the propagated node feature, pipe flow calculation methods in fluid mechanics (such as Manning's formula, Hassen-Williams formula, etc.) are called, and the physical spatial parameters contained in the node features are substituted to calculate the simulated hydraulic data (such as key parameters such as flow rate, pressure, and velocity) corresponding to each pipe segment. The simulated hydraulic data is compared with the pre-acquired actual hydraulic monitoring data of the pipe network (historical effective monitoring data or previous real-time monitoring data samples can be selected) for error calculation (such as mean square error calculation). The error value is used as the optimization target, and the core parameters such as the weights and biases of the graph neural network are iteratively adjusted through the gradient descent algorithm to continuously reduce the deviation between the simulated hydraulic data and the actual hydraulic monitoring data. When the deviation reaches the preset convergence threshold, the iteration stops, and a pipe network simulation model that can accurately simulate the hydraulic operation characteristics of the sewage pipe network is obtained.

[0086] In one embodiment, based on real-time monitoring data and a pipeline network simulation model, an anomaly identification threshold for each pipe segment in the pipeline network simulation model is determined using a particle swarm optimization algorithm to obtain a pipeline network monitoring model, including:

[0087] S21, based on real-time monitoring data and pipeline network simulation model, generates residual time series data for each pipe segment through difference calculation;

[0088] S22, based on residual time series data and combined with preset optimization objectives, multiple candidate threshold schemes are generated by particle swarm optimization algorithm, wherein each candidate threshold scheme contains a set of anomaly identification thresholds corresponding to each pipe segment;

[0089] S23, Based on the preset optimization objective, calculate the fitness score of each candidate threshold scheme through the fitness function;

[0090] S24. Select the optimal anomaly detection threshold scheme from the candidate threshold schemes based on the fitness score.

[0091] S25. Based on the optimal anomaly identification threshold scheme, configure the anomaly identification threshold for each pipe segment in the pipeline network simulation model to obtain the pipeline network monitoring model.

[0092] Specifically, for each pipe segment, the difference between the real-time monitoring data and the corresponding pipe segment simulation data output by the pipe network simulation model is calculated (using absolute error or squared error calculation methods). These difference values ​​are then arranged in chronological order to generate unique residual time-series data for each pipe segment, reflecting the deviation between the actual operating state and the model simulation state of the pipe segment. The preset optimization objective can be "minimizing the weighted sum of the false alarm rate and the missed alarm rate" (where the weights of the false alarm rate and the missed alarm rate can be set according to the actual needs of water management, such as 0.4 and 0.6 respectively). Based on this optimization objective, the particle swarm optimization is initialized. Each particle corresponds to a candidate threshold scheme. The particle dimension is consistent with the number of pipe segments. The value of each dimension is the anomaly identification threshold for the corresponding pipe segment. The threshold range is determined based on the statistical characteristics of the residual time series data (e.g., with 3 times the standard deviation of the residual sequence as the upper limit and 0 as the lower limit). Through the iterative update mechanism of the particle swarm optimization algorithm (in each iteration, the particle adjusts its current position by combining its own historical best position and the global best position), multiple different candidate threshold schemes are generated. The number of iterations can be set to 50-100 times to ensure scheme diversity. Based on the above preset optimization objective, a fitness function is constructed and designed as follows:

[0093]

[0094] Here, α and β are the preset weights for false alarm rate and false negative rate, respectively, and ε is the minimum value of to avoid a denominator of 0. Each candidate threshold scheme is applied to the residual time series data of the corresponding pipe segment, and the false alarm rate (the proportion of samples whose residuals do not exceed the threshold but are judged as abnormal) and false negative rate (the proportion of samples whose residuals exceed the threshold but are not judged as abnormal) under the scheme are calculated. The fitness score of each candidate threshold scheme is calculated by substituting it into the fitness function. The higher the score, the better the anomaly detection effect of the scheme. All candidate threshold schemes are sorted from high to low according to fitness score, and the scheme with the highest score is selected as the optimal anomaly detection threshold scheme. If there are multiple optimal schemes with the same score, the false negative rate can be further compared, and the scheme with the lower false negative rate can be selected as the final optimal scheme. The threshold parameters corresponding to each pipe segment in the optimal anomaly detection threshold scheme are configured into the anomaly detection module of the pipeline network simulation model, so that the model can determine whether there is an anomaly for each pipe segment by comparing its monitoring data with the configured anomaly detection threshold in real time, thus forming a pipeline network monitoring model with anomaly detection capability.

[0095] In one embodiment, based on a pipeline monitoring model and combined with real-time monitoring data, pipeline anomalies are identified, and in response to the identification of pipeline anomalies, potential abnormal pipeline segments are located using a backtracking algorithm, including:

[0096] S31, Based on the anomaly identification threshold of each pipe segment in the pipeline network monitoring model, compare the real-time monitoring data with the anomaly identification threshold of the corresponding pipe segment to identify whether there is anomaly in the pipeline network;

[0097] S32, in response to the detection of pipeline network anomaly, extracts anomaly features from the real-time monitoring data corresponding to the pipeline network anomaly;

[0098] S33, using a bidirectional LSTM algorithm to obtain the time series correlation of abnormal features;

[0099] S34, based on time series correlation and combined with topological structure, uses Dijkstra's algorithm to backtrack and locate the initial abnormal pipe segment;

[0100] S35, based on the initial abnormal pipe segment and combined with the topology, uses a greedy algorithm to expand and screen potential abnormal pipe segments.

[0101] For example, the pre-configured anomaly identification thresholds for each pipe segment in the pipeline monitoring model can be invoked. The real-time monitoring data of each pipe segment is compared with the corresponding threshold. If the real-time monitoring data of a certain pipe segment exceeds or falls below the corresponding threshold for three consecutive sampling cycles (the sampling cycle is set according to the sensor configuration, such as 1 minute / time), then it is determined that there are abnormal signs in that pipe segment. If at least one pipe segment shows such abnormal signs, then an anomaly is identified in the entire pipeline network. In response to the pipeline network anomaly identification result, abnormal features are extracted from the real-time monitoring data corresponding to the anomaly. These features may include the start time of the anomaly, the duration, the deviation of the abnormal value from the threshold, and data changes. Features such as slope and fluctuation frequency during anomalies are quantified into fixed-dimensional feature vectors, forming anomaly feature sets for each anomalous pipe segment. A pre-trained bidirectional LSTM model (a bidirectional long short-term memory network, a recurrent neural network model capable of simultaneously capturing past (forward) and future (backward) contextual information in sequence data; the input dimension is consistent with the anomaly feature vector dimension, the number of hidden layer nodes is set to 64, and the number of iterations is 100) is used to process this anomaly feature set. The forward LSTM layer captures the temporal dependency of the anomaly features from the 10 sampling periods before the anomaly occurs to the time of the anomaly occurrence, while the backward LSTM layer... The TM layer captures the temporal dependencies from the time of an anomaly occurrence to 10 sampling periods after the anomaly. Through bidirectional information fusion, it outputs a feature representation containing complete time-series correlation information, reflecting the evolution of the anomaly over time. Based on this time-series correlation feature, combined with the topology of the sewage pipe network, the order of anomaly occurrence in each anomalous pipe segment is determined according to the temporal correlation (anomalies with earlier timestamps in the feature representation occur first). The Dijkstra algorithm (an algorithm for finding the shortest path from a single source in a weighted graph) is then used to calculate the shortest topology from the pipe segment with the most significant anomaly (largest deviation) to all other anomalous pipe segments. The path (topological distance is measured by the number of nodes connected to the pipe segment) is used to identify the pipe segment that is the starting point of the shortest path and the earliest to show an abnormal time sequence association as the initial abnormal pipe segment. Using the initial abnormal pipe segment as the initial set P, the set C of pipe segments directly connected to the pipe segments in the current P is determined. The objective function value of each pipe segment v∈C is calculated according to the preset greedy algorithm formula. The pipe segment v* with the largest objective function value is selected and added to P. The above process is repeated until the objective function value of the newly added pipe segment is lower than the preset threshold (such as 0.1), at which point the expansion stops. The final set P is the set containing the initial abnormal pipe segment and the closely related potential abnormal pipe segments that may be affected by the abnormality or jointly participate in the abnormality.

[0102] In one embodiment, based on a pipeline monitoring model and considering abnormal pipeline segments, a pipeline scheduling scheme is generated using a dynamic programming algorithm, including:

[0103] S41, Based on potential abnormal pipe sections and simulated hydraulic data, determine the scheduling constraints for each pipe section;

[0104] S42, combining constraints and topology, divide the scheduling phases and define the decision nodes for each scheduling phase;

[0105] S43, construct the stage state space based on decision nodes and constraints;

[0106] S44, input the stage state space into the dynamic programming algorithm to calculate the local scheduling scheme of the decision nodes in each scheduling stage;

[0107] S45, combining the topology, backtracking the local scheduling schemes of each scheduling stage to obtain the global scheduling path;

[0108] S46, Correct conflicting nodes in the global scheduling path using a neighborhood search algorithm to generate an initial scheduling scheme;

[0109] S47. Based on real-time monitoring data, the parameters of the initial scheduling scheme are optimized to obtain the pipeline scheduling scheme.

[0110] Specifically, based on the operational limitations of potentially abnormal pipe sections (such as the maximum allowable flow rate of damaged pipe sections and the upper limit of flow velocity in blocked pipe sections) and the simulated hydraulic data output by the pipe network simulation model (including the design flow rate, rated pressure, critical flow velocity, etc. of each pipe section), the scheduling constraints for each pipe section are determined. Hard constraints include a flow rate not exceeding 120% of the design flow rate and a pressure not lower than the minimum safe pressure of 0.2 MPa. Soft constraints include energy consumption not exceeding 110% of the rated energy consumption and a scheduling response delay not exceeding 5 minutes. The priority of each constraint is as follows: hard constraints take precedence over soft constraints. Combining the above constraints with the topology of the sewage pipe network, the process is divided into three phases based on time: emergency response phase (0-30 minutes, focusing on emergency pressure relief or flow restriction of abnormal pipe sections), optimization and adjustment phase (30-60 minutes, achieving balanced flow distribution in the pipe network), and stable operation phase (after 60 minutes, maintaining normal high-speed operation of the pipe network). The system is divided into three scheduling stages: efficient operation, and inefficient operation. Key equipment (such as pump stations, regulating valves, and diversion valves) and important pipeline connection nodes with scheduling and control capabilities in each stage are defined as decision nodes. Each decision node corresponds to a specific control object (such as pump station P1 and valve V2). Based on the decision nodes and corresponding constraints of each scheduling stage, a stage state space is constructed. Each state in the state space contains the core operating parameter values ​​of the decision node (such as valve opening 0-100%, pump station operating frequency 30-50Hz), and each state satisfies the constraints of the current stage. For example, in the emergency response stage, the valve opening state corresponding to the abnormal pipeline segment needs to be limited to 0-50% to achieve flow restriction. The constructed stage state space is input into a preset dynamic programming algorithm, with the objective function set as "minimize pipeline network operating energy consumption + maximize anomaly mitigation efficiency" (the weights of which are set to 0.4 and 0, respectively).6) The state transition equation is constructed based on the hydraulic characteristics of the pipeline network (such as the linear relationship between flow rate and valve opening). It iteratively calculates the local scheduling schemes of each decision node within each scheduling stage, i.e., the optimal operating parameters of each decision node in the current stage (e.g., pump station P1 operating frequency 45Hz, valve V2 opening 30% in the emergency stage). Combining the connection relationships between pipe segments and decision nodes in the pipeline network topology, starting from the optimal local scheduling scheme of the last scheduling stage, it backtracks to the optimal state of each decision node in the preceding stage, forming a global scheduling path that runs through the three stages of emergency response, optimization adjustment, and stable operation. The scheduling schemes of each stage are mutually matched (e.g., the flow restriction scheme in the emergency stage is consistent with the flow diversion scheme in the adjustment stage). If there are conflicting nodes in the global scheduling path (e.g., the scheduling schemes of two adjacent decision nodes cause the pressure in the intermediate pipe segment to exceed the limit), a neighborhood search algorithm can be used to correct the conflicting nodes. Centered on the current state of the conflicting node, a feasible state satisfying the constraints is searched within its neighborhood states (e.g., valve opening ±10%, pump station frequency ±5Hz). This alternative state replaces the original conflicting state, eliminating all scheduling conflicts and generating an initial scheduling scheme. The latest real-time monitoring data of the pipeline network (e.g., actual flow rate, pressure, and velocity of each pipe segment) is acquired and compared with the simulated data corresponding to the initial scheduling scheme. Deviation values ​​(e.g., flow deviation rate, pressure deviation) are calculated. With the goal of minimizing the deviation, the operating parameters of the decision nodes in the initial scheduling scheme are iteratively adjusted using gradient descent (e.g., fine-tuning valve opening from 30% to 32%, pump station frequency from 45Hz to 43Hz). Optimization stops when the deviation value is lower than a preset threshold (e.g., flow deviation rate ≤5%, pressure deviation ≤0.05MPa). This yields a pipeline network scheduling scheme that closely matches the actual operating state of the pipeline network, satisfies the constraints, and achieves the optimal optimization objective.

[0111] In one embodiment, S51, based on the initial abnormal pipe segment and combined with the topology, potential abnormal pipe segments are obtained through greedy algorithm expansion screening. The greedy algorithm expansion screening is implemented through the following formula:

[0112]

[0113] in, Let C be the pipe segment selected in the current greedy step, and let C be the pipe segment adjacent to the current potentially abnormal pipe segment P. This refers to the set of pipe segments directly connected to pipe segment v. Let U be the topological distance between pipe segments u and v, and P represent the potential abnormal pipe segment in the current step. P is initialized to the initial abnormal pipe segment.

[0114] For example, the initial abnormal pipe segments can be used as the initial set P. Based on the topology of the sewage pipe network, all pipe segments that have a direct connection relationship with any pipe segment in the current set P (i.e., there is an edge connecting them in the topology) are selected to form a set C of pipe segments adjacent to the current P (excluding pipe segments already included in P). For each candidate pipe segment v in set C, the set of pipe segments directly connected to it is determined. (That is, all pipe segments in the topology that are directly connected to v through edges), and then from The subset of pipe segments that simultaneously belong to the current set of potentially abnormal pipe segments P is selected (i.e., ), calculate the topological distance between each pipe segment u and the candidate pipe segment v within this subset. (The number of pipe segments included in the shortest connection path between pipe segments u and v, when directly connected) =1, when connected with a gap of 1 pipe segment =2, and so on), and calculate the corresponding value for each u. The objective function value of candidate pipe segment v is obtained by summing these values. According to the greedy algorithm formula, the candidate pipe segment with the largest objective function value is selected from set C as the selected pipe segment v* in the current greedy step (if there are multiple candidate pipe segments with the same objective function value, the pipe segment with the closest topological distance to the initial abnormal pipe segment is selected first), and v is added to the potential abnormal pipe segment set P to complete one greedy expansion. The above process is repeated to redetermine the pipe segment set C adjacent to the updated set P (excluding pipe segments already in P), and the objective function value of each candidate pipe segment in the new set C is calculated again and the optimal pipe segment is selected to be added to P. The iteration stops when the objective function value of the newly added pipe segment v is lower than the preset threshold (e.g., 0.1, which is preset according to the pipe network topology complexity and abnormal propagation characteristics), or when the number of pipe segments in set P reaches the preset upper limit (e.g., 3-5 times the number of the initial abnormal pipe segments to avoid over-expansion). The resulting set P is the set containing the initial abnormal pipe segment and the closely related potential abnormal pipe segments that may be affected by the abnormal or participate in the abnormal propagation.

[0115] In one embodiment, real-time monitoring data, pipeline monitoring models, potentially abnormal pipeline segments, and pipeline scheduling schemes are integrated, and visualization results are generated through visualization methods, including:

[0116] S61, based on real-time monitoring data, pipeline monitoring model and potential abnormal pipe sections, extracts abnormal feature data and potential abnormal pipe section location information;

[0117] S62 integrates abnormal feature data, potential abnormal pipe segment location information, and pipeline scheduling schemes to construct a visual basic dataset;

[0118] S63, define the visualization hierarchy of each element within the basic visualization dataset;

[0119] S64, combined with the topology, uses a coordinate mapping algorithm to determine the spatial display position of each visualization level;

[0120] S65, based on spatial display location and visualization hierarchy, integrates various elements in the visualization base dataset into the spatial display interface through dynamic annotation to form a visualization result.

[0121] Specifically, based on abnormal data items exceeding the anomaly identification threshold in real-time monitoring data, the anomaly judgment criteria output by the pipeline monitoring model (such as threshold range and judgment logic), and the identification information of potential abnormal pipeline segments, abnormal feature data (quantitative indicators such as anomaly type, anomaly duration, data deviation amplitude, and anomaly frequency) and potential abnormal pipeline segment location information (including the unique identifier ID of the pipeline segment, the corresponding geographical coordinates latitude and longitude, and the node connection relationship in the topology) can be extracted. A data association and fusion algorithm is then used, with the unique identifier ID of the pipeline segment as the association key, to link the abnormal feature data, potential abnormal pipeline segment location information, and the pipeline network scheduling scheme (including scheduling objects, control parameters, and execution sequence). The system performs field matching and redundant data removal for data from various sources (including path planning, etc.), integrating data from different sources into a unified format for visualization. Each data record contains four core fields: pipe segment ID, location information, anomaly characteristics, and scheduling instructions. The visualization is divided into layers based on "data importance + display priority": a bottom-level pipeline topology foundation layer (displaying the overall pipeline connection structure and pipe segment distribution), a middle-level real-time status layer (displaying real-time monitoring data and operating status of each pipe segment), an upper-level anomaly labeling layer (displaying the location and characteristics of potentially abnormal pipe segments), and a top-level scheduling scheme layer (displaying scheduling paths and control parameters). The priority of each layer increases from bottom to top. Information is prioritized for visualization. Combining the topology of the sewage pipe network, coordinate mapping algorithms (such as UTM projection transformation algorithms or pixel coordinate mapping formulas) are used to convert the geographical coordinates (latitude and longitude) or topological relative positions of potentially abnormal pipe segments into pixel coordinates on the visualization interface. Based on the priority of each visualization level, the spatial stacking order of elements in each layer is determined (the pixel coordinate range of lower-level elements does not obscure the core information of upper-level elements), clearly defining the specific display positions of elements such as pipe network topology, anomaly markings, and scheduling paths on the interface. Based on the determined spatial display positions and visualization levels, dynamic annotation technology is used to render the interface of each element. The basic layer of the pipe network topology uses gray solid lines to draw pipe segments and black... The system uses dots to mark nodes, and the real-time status layer displays data such as flow and pressure of each pipe section synchronously with dynamic numerical labels (the data update frequency is consistent with the monitoring data sampling cycle, set to 1 minute / time). The anomaly labeling layer highlights potentially abnormal pipe sections with flashing red borders, and clicking on them displays the abnormal feature data in a pop-up window. The scheduling scheme layer marks the scheduling path with solid blue lines with arrows, and the arrow direction corresponds to the water flow control direction. The control parameters are marked with yellow floating labels at the corresponding scheduling nodes (such as pump stations and valves). All hierarchical elements are integrated into an interactive spatial display interface that supports zooming, panning, and click-to-query, forming a visualization result that can intuitively present the pipeline network operation status, anomaly distribution, and intelligent scheduling scheme.

[0122] The aforementioned sewage pipe network visualization method provides comprehensive data support for pipe network monitoring and scheduling by acquiring the topology, physical spatial parameters, and real-time monitoring data of the sewage pipe network. Based on the topology, it constructs graph structure data and assigns physical spatial parameter feature vectors. Simulated hydraulic data is generated through graph neural network feature encoding, multi-layer graph convolutional propagation, and fluid dynamics methods. The model parameters are optimized using gradient descent algorithm to construct a pipe network simulation model that captures the global spatial correlation and fluid characteristics of the pipe network, overcoming the limitation of traditional methods lacking global topological correlation. The method utilizes particle swarm optimization to process the residual time-series data between real-time monitoring and simulation data, generating and selecting the optimal anomaly identification threshold scheme. After configuration, a pipe network monitoring model is obtained, improving the accuracy and adaptability of anomaly identification and solving the false alarm and missed alarm problems of traditional fixed threshold alarms. When a pipe network anomaly is identified, it extracts anomaly features and performs bidirectional... LSTM is used to acquire temporal correlations, and Dijkstra's algorithm is used to backtrack and locate initial abnormal pipe segments. A greedy algorithm is then used to expand and filter potential abnormal pipe segments, achieving rapid and accurate location of abnormal segments and avoiding the drawbacks of low efficiency and slow location in traditional manual inspections. Based on potential abnormal pipe segments and simulated hydraulic data, scheduling constraints are determined. A dynamic programming algorithm is used to divide scheduling stages, construct a state space, calculate local scheduling schemes, and backtrack and integrate them. Neighborhood search algorithms are combined to correct conflicts and optimize real-time data, generating an intelligent pipe network scheduling scheme. This solves the problem of response delays and resource waste caused by the lack of effective scheduling strategies in traditional methods. Multi-source data is integrated, and visualization results are generated through hierarchical division, coordinate mapping, and dynamic annotation, presenting the pipe network status, anomaly distribution, and scheduling schemes. This provides efficient decision support for water management and improves the intelligence level of sewage pipe network operation monitoring and emergency response efficiency.

[0123] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0124] Based on the same inventive concept, this application also provides a sewage network visualization device for implementing the above-described sewage network visualization method. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations of one or more embodiments of the sewage network visualization device provided below can be found in the limitations of the sewage network visualization method described above, and will not be repeated here.

[0125] In one exemplary embodiment, such as Figure 2 As shown, a sewage pipe network visualization device is provided, comprising:

[0126] Pipeline data acquisition module 101 is used to acquire the topology, physical spatial parameters and real-time monitoring data of the sewage pipeline network;

[0127] The simulation model construction module 102 is used to construct a pipeline network simulation model based on topology and physical space parameters using a graph neural network.

[0128] The monitoring model generation module 103 is used to determine the anomaly identification threshold of each pipe segment in the pipeline simulation model based on real-time monitoring data and pipeline simulation model, and obtain the pipeline monitoring model by using the particle swarm optimization algorithm.

[0129] The anomaly identification and location module 104 is used to identify pipeline anomalies based on the pipeline monitoring model and combined with real-time monitoring data, and to locate potential abnormal pipe sections through a backtracking algorithm in response to the identification of pipeline anomalies.

[0130] The scheduling scheme generation module 105 is used to generate a pipeline scheduling scheme based on the pipeline monitoring model and combined with abnormal pipeline sections through a dynamic programming algorithm.

[0131] The visualization result generation module 106 is used to integrate real-time monitoring data, pipeline monitoring models, potential abnormal pipeline sections, and pipeline scheduling schemes, and generate visualization results through visualization methods.

[0132] In one embodiment, the simulation model building module 102 is further configured to:

[0133] Based on the topology, a graph structure data of the sewage pipe network is constructed, in which nodes represent pipe segments and edges represent the connection relationships between pipe segments.

[0134] Based on physical space parameters, feature vectors are assigned to each node of the graph structure data to form graph structure data with node features;

[0135] Graph structure data with node features is input into a graph neural network, and node-encoded features are generated through feature encoding.

[0136] Extract the adjacency matrix from the graph structure data;

[0137] Based on node encoding features and combined with the adjacency matrix, multi-layer graph convolution propagation is used to generate propagated node features;

[0138] Based on the characteristics of the nodes after propagation, simulated hydraulic data for each pipe segment are generated using fluid dynamics methods.

[0139] Based on simulated hydraulic data, the parameters of the graph neural network are adjusted using the gradient descent algorithm to obtain a pipeline network simulation model.

[0140] In one embodiment, the monitoring model generation module 103 is further configured to:

[0141] Based on real-time monitoring data and pipeline network simulation models, residual time-series data for each pipeline segment are generated through difference calculation;

[0142] Based on residual time series data and combined with preset optimization objectives, multiple candidate threshold schemes are generated by particle swarm optimization algorithm. Each candidate threshold scheme contains a set of anomaly identification thresholds corresponding to each pipe segment.

[0143] Based on the preset optimization objective, the fitness score of each candidate threshold scheme is calculated using a fitness function;

[0144] Based on the fitness score, the optimal anomaly detection threshold scheme is selected from the candidate threshold schemes;

[0145] Based on the optimal anomaly identification threshold scheme, the anomaly identification threshold for each pipe segment in the pipeline network simulation model is configured to obtain the pipeline network monitoring model.

[0146] In one embodiment, the anomaly identification and location module 104 is further configured to:

[0147] Based on the anomaly identification threshold of each pipe segment in the pipeline network monitoring model, the real-time monitoring data is compared with the anomaly identification threshold of the corresponding pipe segment to identify whether there are anomalies in the pipeline network.

[0148] In response to the identification of pipeline network anomalies, abnormal features are extracted from the real-time monitoring data corresponding to the pipeline network anomalies;

[0149] A bidirectional LSTM algorithm is used to obtain the time series correlation of abnormal features;

[0150] Based on time series correlation and combined with topological structure, the initial abnormal pipe segment is located by backtracking using the Dijkstra algorithm.

[0151] Based on the initial abnormal pipe segments and combined with the topology, potential abnormal pipe segments are obtained through a greedy algorithm for further screening.

[0152] In one embodiment, the scheduling scheme generation module 105 is further configured to:

[0153] Based on potential abnormal pipe sections and simulated hydraulic data, the scheduling constraints for each pipe section are determined.

[0154] Based on constraints and topology, the scheduling phases are divided and the decision nodes for each scheduling phase are defined.

[0155] Construct a stage state space based on decision nodes and constraints;

[0156] The stage state space is input into the dynamic programming algorithm to calculate the local scheduling scheme of the decision nodes in each scheduling stage;

[0157] By combining the topology, we can backtrack the local scheduling schemes of each scheduling stage to obtain the global scheduling path;

[0158] The conflicting nodes in the global scheduling path are corrected by a neighborhood search algorithm to generate an initial scheduling scheme.

[0159] Based on real-time monitoring data, the parameters of the initial scheduling scheme are optimized to obtain the pipeline scheduling scheme.

[0160] In one embodiment, the anomaly identification and location module 104 uses the following formula to implement a greedy algorithm for extended screening based on the initial abnormal pipe segment and the topology structure, thereby obtaining potential abnormal pipe segments through greedy algorithm-based extended screening:

[0161]

[0162] in, Let C be the pipe segment selected in the current greedy step, and let C be the pipe segment adjacent to the current potentially abnormal pipe segment P. This refers to the set of pipe segments directly connected to pipe segment v. Let U be the topological distance between pipe segments u and v, and P represent the potential abnormal pipe segment in the current step. P is initialized to the initial abnormal pipe segment.

[0163] In one embodiment, the visualization result generation module 106 is further configured to:

[0164] Based on real-time monitoring data, pipeline monitoring models, and potential abnormal pipe sections, extract abnormal feature data and location information of potential abnormal pipe sections;

[0165] By integrating abnormal feature data, potential abnormal pipe segment location information, and pipeline network scheduling schemes, a visual basic dataset is constructed.

[0166] Define the visualization hierarchy of each element within the basic visualization dataset;

[0167] Based on the topology, the spatial display position of each visualization level is determined through a coordinate mapping algorithm;

[0168] Based on the spatial display location and visualization hierarchy, the elements in the basic visualization dataset are integrated into the spatial display interface through dynamic annotation to form a visualization result.

[0169] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the aforementioned wastewater network visualization method.

[0170] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0171] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0172] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A sewer network visualisation method, characterised in that, The method comprises: acquiring the topological structure, physical space parameters and real-time monitoring data of the sewage pipe network; based on the topological structure and the physical space parameters, a pipe network simulation model is constructed using a graph neural network; based on the real-time monitoring data and the pipe network simulation model, the abnormal identification threshold of each pipe segment in the pipe network simulation model is determined by a particle swarm optimization algorithm to obtain a pipe network monitoring model; based on the pipe network monitoring model, in combination with the real-time monitoring data, pipe network anomaly identification is performed, and in response to identifying a pipe network anomaly, a backtracking algorithm is used to locate a potential abnormal pipe segment; based on the pipe network monitoring model, in combination with the abnormal pipe segment, a pipe network scheduling scheme is generated by a dynamic programming algorithm; the real-time monitoring data, the pipe network monitoring model, the potential abnormal pipe segment and the pipe network scheduling scheme are integrated to generate a visual result by a visualization method.

2. The method of claim 1, wherein, The method comprises: based on the topological structure, a graph structure data of the sewage pipe network is constructed, wherein the graph structure data represents pipe segments as nodes and the connection relationship between pipe segments as edges; based on the physical space parameters, a feature vector is assigned to each node of the graph structure data to form a graph structure data with node features; the graph structure data with node features is input into a graph neural network to generate node encoding features through feature encoding; the adjacency matrix of the graph structure data is extracted; based on the node encoding features, in combination with the adjacency matrix, a multi-layer graph convolution propagation is performed to generate propagated node features; based on the propagated node features, simulation hydraulic data of each pipe segment is generated by a fluid mechanics method; based on the simulation hydraulic data, the parameters of the graph neural network are adjusted by a gradient descent algorithm to obtain the pipe network simulation model.

3. The method of claim 2, wherein, The method comprises: based on the real-time monitoring data and the pipe network simulation model, residual time series data of each pipe segment is generated by difference calculation; based on the residual time series data, in combination with a preset optimization target, a plurality of candidate threshold schemes are generated by a particle swarm optimization algorithm, wherein each candidate threshold scheme includes a set of abnormal identification thresholds corresponding to each pipe segment; based on the preset optimization target, the fitness score of each candidate threshold scheme is calculated by a fitness function; based on the fitness score, the optimal abnormal identification threshold scheme is selected from the candidate threshold schemes; based on the optimal abnormal identification threshold scheme, the abnormal identification threshold of each pipe segment in the pipe network simulation model is configured to obtain the pipe network monitoring model.

4. The method of claim 3, wherein, The method comprises: based on the abnormal identification threshold of each pipe segment of the pipe network monitoring model, the real-time monitoring data is compared with the abnormal identification threshold of the corresponding pipe segment to identify whether there is an anomaly in the pipe network; extracting an abnormal feature from real-time monitoring data corresponding to the pipe network anomaly in response to identifying the pipe network anomaly; acquiring time sequence correlation of the abnormal feature by using a bidirectional LSTM algorithm; locating an initial abnormal pipe section by backtracking based on the time sequence correlation and the topological structure through a Dijkstra algorithm; obtaining the potential abnormal pipe section by expanding and screening based on the initial abnormal pipe section and the topological structure through a greedy algorithm.

5. The method of claim 2, wherein, generating a pipe network scheduling scheme by dynamic programming algorithm based on the pipe network monitoring model, the abnormal pipe section, including: determining scheduling constraint conditions of each pipe section based on the potential abnormal pipe section and the simulated hydraulic data; dividing scheduling stages and defining decision nodes of each scheduling stage in combination with the constraint conditions and the topological structure; constructing a stage state space based on the decision nodes and the constraint conditions; inputting the stage state space into a dynamic programming algorithm to calculate local scheduling schemes of the decision nodes of each scheduling stage; obtaining a global scheduling path by backtracking the local scheduling schemes of each scheduling stage in combination with the topological structure; generating an initial scheduling scheme by correcting conflict nodes in the global scheduling path through a neighborhood search algorithm; optimizing parameters of the initial scheduling scheme based on the real-time monitoring data to obtain the pipe network scheduling scheme.

6. The method of claim 4, wherein, The potential abnormal pipe section is obtained by expanding and screening based on the initial abnormal pipe section and the topological structure through a greedy algorithm, which is realized by the following formula: wherein, is the pipe segment selected in the current greedy step, C is the pipe segment adjacent to the current potential abnormal pipe segment P, is the set of pipe segments directly connected to pipe segment v, is the topological distance between pipe segments u and v, P denotes the potential abnormal pipe segment of the current step, P is initialized as the initial abnormal pipe segment.

7. The method of claim 1, wherein, The real-time monitoring data, the pipe network monitoring model, the potential abnormal pipe section and the pipe network scheduling scheme are integrated to generate a visual result by a visualization method, including: extracting abnormal feature data and potential abnormal pipe section location information based on the real-time monitoring data, the pipe network monitoring model and the potential abnormal pipe section; fusing the abnormal feature data, the potential abnormal pipe section location information and the pipe network scheduling scheme to construct a visualization basic data set; dividing visualization levels of each element in the visualization basic data set; determining spatial display positions of each visualization level by a coordinate mapping algorithm in combination with the topological structure; integrating each element in the visualization basic data set into a spatial display interface by dynamic labeling based on the spatial display positions and the visualization levels to form the visualization result.

8. A sewer network visualisation apparatus, characterised in that, The device comprises: a pipe network data acquisition module for acquiring a topological structure, physical space parameters and real-time monitoring data of a sewage pipe network; a simulation model construction module for constructing a pipe network simulation model by using a graph neural network based on the topological structure and the physical space parameters; a monitoring model generation module for determining an abnormality identification threshold of each pipe section in the pipe network simulation model by a particle swarm optimization algorithm based on the real-time monitoring data and the pipe network simulation model to obtain a pipe network monitoring model; an abnormality identification and positioning module for identifying a pipe network anomaly based on the pipe network monitoring model in combination with the real-time monitoring data, and locating a potential abnormal pipe section by a backtracking algorithm in response to identifying the pipe network anomaly. a scheduling scheme generation module, configured to generate a pipe network scheduling scheme by a dynamic programming algorithm based on the pipe network monitoring model and in combination with the abnormal pipe section; a visual result generation module, configured to integrate the real-time monitoring data, the pipe network monitoring model, the potential abnormal pipe section and the pipe network scheduling scheme, and generate a visual result by a visual method. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The processor implements the steps of the method of any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by the processor, implements the steps of the method of any one of claims 1 to 7.

Images