Municipal sewage pipe network leakage detection system and method

Abstract

The invention relates to the technical field of town sewage pipe network detection, and discloses a town sewage pipe network leakage detection system and method. The system comprises a data acquisition module which uses a multi-source sensor to acquire real-time operation data such as pipe network pressure, flow and the like; the feature extraction module extracts spatio-temporal features based on the multi-scale convolutional neural network, and generates a pipe network state feature matrix; the anomaly detection module inputs the feature matrix into a pre-training model and marks a potential leakage area; the optimization analysis module constructs a multi-constraint dynamic optimization model, and pipe network pressure parameters are optimized by using an adaptive particle swarm algorithm; the hierarchical execution module generates a global regulation and control sequence, dynamically matches local pressure parameters and adjusts the valve opening and the pump station power through a decision layer, a region coordination layer and an execution layer. The system and the method are accurate in detection and reasonable in regulation and control optimization, leakage risks can be effectively reduced, the operation management level of a pipe network is improved, and water resource waste and environmental pollution are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of urban sewage pipe network detection, and in particular to an urban sewage pipe network leakage detection system and method. Background Art

[0002] In the urban sewage pipe network system, pipe leakage has always been a major problem that plagues the management and maintenance of urban infrastructure. With the rapid advancement of urbanization, the scale of urban sewage pipe networks continues to expand, the layout becomes more complex, and factors such as pipe aging, construction defects, and external loads have led to frequent leakage.

[0003] The traditional manual inspection method is inefficient and inaccurate, making it difficult to cover large-scale pipe network systems, and is easily affected by the experience and subjective factors of the inspectors. For example, manual inspection can only judge whether there is leakage by simple methods such as observing whether there is sewage seepage on the ground and listening to whether there is abnormal sound in the pipe. It is difficult to find leakage points of pipes that are deeply buried and highly concealed in time. Although some existing detection technologies, such as acoustic detection and pressure detection, have improved the accuracy of detection to a certain extent, they also have problems such as limited detection range and poor adaptability to complex pipe networks. These technologies can often only detect local areas of the pipeline, and it is difficult to fully grasp the leakage of the entire pipe network, and cannot meet the needs of efficient and accurate management of modern urban sewage pipe networks. Therefore, it is urgent to develop an efficient, accurate and adaptable urban sewage pipe network leakage detection system and method. Summary of the invention

[0004] The purpose of the present invention is to provide a system and method for detecting leakage of urban sewage pipe networks to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solution: a leakage detection system for urban sewage pipe network, the system comprising: Data acquisition module: used to collect real-time operation data of the pipe network through multi-source sensors, including pressure, flow, acoustic wave signals and water quality parameters; Feature extraction module: extracting spatiotemporal features of the real-time operation data based on a multi-scale convolutional neural network to generate a pipeline network state feature matrix; Anomaly detection module: input the pipeline network state feature matrix into the pre-trained spatiotemporal anomaly detection model to generate potential leakage area markers; Optimization analysis module: construct a multi-constraint dynamic optimization model based on the leakage area mark, the model takes pressure balance adjustment and leakage positioning accuracy as optimization goals, and uses an adaptive particle swarm algorithm to globally optimize the pipe network pressure parameters; Hierarchical Execution Module: Based on the output of the dynamic optimization model, the optimal regulation strategy is realized, and the distributed response of instructions is achieved through a hierarchical execution architecture, which includes a decision-making layer, a regional coordination layer, and an execution layer. The decision-making layer generates a global pressure regulation sequence, the regional coordination layer dynamically matches local pressure parameters using a sliding window optimization algorithm, and the execution layer adjusts the valve opening and pump station power based on a fuzzy proportional integral control algorithm.

[0006] Preferably, the acquisition of real-time pipeline network operation data through multi-source sensors includes: The multi-source sensors include pressure transmitters, electromagnetic flowmeters, acoustic wave sensors, and dissolved oxygen sensors; Perform spatio-temporal alignment on the pressure transmitter data and electromagnetic flowmeter data to construct a pipeline network pressure-flow distribution map; perform wavelet packet decomposition on the acoustic wave sensor data to generate a high-frequency vibration feature sequence; Construct a dual-branch feature fusion network. The first branch uses a three-dimensional convolutional network to extract the spatial correlation features of the pressure-flow distribution map, and the second branch uses a temporal convolutional network to extract the time-varying features of the high-frequency vibration feature sequence; Fuse the spatial correlation features and time-varying features through a cross-modal gating mechanism to generate a joint feature tensor; perform dynamic mode modeling on the joint feature tensor based on a bidirectional long short-term memory network, and output a pipeline network state feature matrix including pressure fluctuations, vibration anomalies, and water quality mutations.

[0007] Preferably, the spatio-temporal anomaly detection model adopts a multi-level graph neural network structure and generates leakage area markers based on a dynamic graph attention mechanism; the structure includes: Construct a pipeline network-environment interaction graph. The nodes in the graph include pipeline nodes, pump station nodes, valve nodes, and environmental monitoring nodes, and the node attributes include pressure values, flow values, and acoustic wave energy; Adopt a dual-path attention mechanism. The first path calculates the interaction weights between pipeline nodes and adjacent nodes through a spatial graph aggregation layer, and the second path performs feature screening on historical operation data through a temporal graph pooling layer; Iteratively update the node attributes based on a multi-head graph attention module. Each attention head fuses the node state and environmental monitoring parameters; stabilize the training process through a skip connection and layer normalization mechanism, and finally output leakage area markers including pipeline network topological constraints.

[0008] Preferably, the adaptive particle swarm optimization algorithm integrates an inertia weight adjustment and a constraint repair strategy, including: Model the pressure optimization problem as a multi-constraint non-linear programming problem, and the decision variables include discrete valve state variables and continuous pump station power variables; Initialize the particle swarm and calculate the Pareto front candidate solutions, and adopt an inertia weight adaptive mechanism to adjust the particle search range according to the real-time pipe network load; In the optimization stage, divide the global problem into multiple sub-group optimization problems based on the objective decomposition technology; in the constraint repair stage, introduce a relaxation factor to balance the conflicting constraints between sub-groups; Adopt a parallel evolution strategy to iteratively optimize the sub-groups, update the local optimal solution in each iteration, and synchronize the optimal solution through global information sharing.

[0009] Preferably, the regional coordination layer uses a sliding window optimization algorithm to dynamically match the local pressure parameters, including: Construct a time-varying pipe network pressure propagation model, discretize the hydrodynamic equation into a state transition equation, and the equation includes a pressure attenuation gradient, a flow rate mutation term, and a valve response delay term; Design a window optimization objective function, including a pressure fluctuation suppression term, an energy consumption minimization term, and a leakage location error correction term.

[0010] Preferably, the execution layer realizes the adjustment of the valve opening and the pump station power based on the fuzzy proportional integral control algorithm, including: Design a multi-level fuzzy inference system to map the pressure error and the flow rate error to the fuzzy rule triggering degree; construct an adaptive proportional integral parameter adjustment mechanism to dynamically adjust the slope of the fuzzy membership function according to the error accumulation.

[0011] Preferably, the three-dimensional convolutional network adopts a multi-scale dilated convolutional structure to accelerate feature extraction, including: Divide the pipe network pressure-flow distribution map into multi-level grids, and each grid cell stores the pressure gradient and the flow rate change rate statistics; In the encoding stage, use a dilated convolutional kernel to expand the spatial perception range, and in the decoding stage, restore the local details through a transposed convolutional layer; Introduce a spatial attention module to allocate regional weights to the feature map.

[0012] Preferably, the spatial graph aggregation layer adopts a relative distance encoding mechanism, including: Define the relative distance vector between the pipe node and the adjacent node, including the hydraulic gradient, the pipe material attenuation coefficient, and the connection direction; Convert the relative distance vector into the offset parameter of the graph convolutional kernel through a non-linear mapping layer; Superimpose the offset parameter on the standard graph convolution operation.

[0013] Preferably, the inertia weight adaptive mechanism is realized based on an online feedback strategy, including: Collect the particle fitness value and the constraint violation degree in the historical optimization process as training data; Construct the mapping relationship between the fitting weight coefficients of the extreme learning machine network and the pipe network state; Update the network parameters through the online recursive least squares method and adjust the particle swarm search step size in real time; When a sudden change in the pipe network pressure is detected, trigger the rapid weight reset operation.

[0014] Preferably, the present invention further includes a method for detecting leakage in urban sewage pipe networks, which is applied to the above-mentioned urban sewage pipe network leakage detection system. The method includes the following steps: Step 1: Use a multi-source sensor through the data acquisition module to collect real-time operation data of the pipe network. The real-time operation data includes pressure, flow rate, acoustic signal, and water quality parameters; Step 2: Use the feature extraction module to extract spatio-temporal features from the real-time operation data collected in Step 1 based on a multi-scale convolutional neural network to generate a pipe network state feature matrix; Step 3: Input the pipe network state feature matrix generated in Step 2 into a pre-trained spatio-temporal anomaly detection model, and generate potential leakage area markers through the anomaly detection module; Step 4: The optimization analysis module constructs a multi-constraint dynamic optimization model according to the leakage area markers obtained in Step 3. The model takes pressure balance adjustment and leakage location accuracy as optimization objectives, and uses an adaptive particle swarm algorithm to globally optimize the pipe network pressure parameters; Step 5: Based on the optimal control strategy output by the dynamic optimization model in Step 4, the hierarchical execution architecture of the hierarchical execution module realizes the distributed response of instructions. Among them, the decision-making layer generates a global pressure control sequence, the regional coordination layer uses a sliding window optimization algorithm to dynamically match local pressure parameters, and the execution layer realizes the adjustment of valve opening and pump station power based on the fuzzy proportional integral control algorithm.

[0015] Compared with the prior art, the beneficial effects of the present invention are: In terms of detection accuracy, the data acquisition module uses multi-source sensors, such as pressure transmitters, electromagnetic flowmeters, acoustic sensors, and dissolved oxygen sensors, to collect rich real-time operation data of the pipe network, covering pressure, flow rate, acoustic signal, and water quality parameters. The data is processed through technologies such as spatio-temporal alignment and wavelet packet decomposition, and with the help of a dual-branch feature fusion network and a bidirectional long short-term memory network, the data features are deeply mined to generate a feature matrix that accurately reflects the pipe network state. The anomaly detection module adopts a multi-level graph neural network structure and a dynamic graph attention mechanism, fully considering the pipe network topology structure and environmental factors, and can accurately locate potential leakage areas. Compared with traditional detection methods, it greatly improves the accuracy and reliability of detection, and reduces the situations of missed detection and false detection.

[0016] In terms of optimization control, the optimization analysis module constructs a multi-constraint dynamic optimization model, with pressure balance adjustment and leakage location accuracy as the optimization objectives, and uses the adaptive particle swarm optimization algorithm to globally optimize the pipe network pressure parameters. This algorithm integrates the inertia weight adjustment and constraint repair strategies, dynamically adjusts the particle search range according to the real-time pipe network load, effectively solves the multi-constraint nonlinear programming problem in pressure optimization, realizes the precise control of the pipe network pressure, ensures the stability of the pipe network operation, and reduces the leakage risk caused by abnormal pressure.

[0017] The hierarchical execution architecture of the hierarchical execution module has efficient instruction response capabilities. The decision-making layer generates the global pressure control sequence to overall control the pipe network operation; the regional coordination layer constructs a time-varying pipe network pressure propagation model, uses the sliding window optimization algorithm, comprehensively considers pressure fluctuation suppression, energy consumption minimization and leakage location error correction, dynamically matches the local pressure parameters, and improves the flexibility and accuracy of local control; the execution layer is based on the fuzzy proportional integral control algorithm, and through designing a multi-level fuzzy inference system and an adaptive proportional integral parameter adjustment mechanism, accurately adjusts the valve opening and pump station power according to the pressure and flow errors, realizes the refined control of the pipe network equipment, effectively reduces the energy consumption, and improves the energy utilization efficiency.

[0018] From the perspective of the overall system operation, the invention realizes the full-process intelligent management of urban sewage pipe networks from data collection, feature extraction, anomaly detection to optimization control. It can not only timely detect leakage hazards, but also reduce the leakage incidence rate through optimization control measures, reduce water resource waste and environmental pollution, reduce the maintenance cost of urban sewage pipe networks, improve the intelligent level of urban infrastructure management, ensure the safe, stable and efficient operation of urban sewage pipe networks, and provide strong support for the sustainable development of the city. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is the working principle diagram of a leakage detection system for an urban sewage pipe network according to the present invention; Figure 2 It is the working principle diagram of data collection and feature extraction; Figure 3 It is the working principle diagram of the adaptive particle swarm optimization algorithm; Figure 4 It is the working principle diagram of the execution layer. DETAILED DESCRIPTION OF THE INVENTION

[0020] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0021] Please refer to Figures 1 - 4 , the present invention provides a technical solution: a leakage detection system for urban sewage pipe networks, and the system includes: Data acquisition module: This module uses multi-source sensors to collect real-time operation data of the pipe network. These data cover pressure, flow rate, acoustic signals, and water quality parameters. The multi-source sensors can specifically adopt devices such as pressure transmitters, electromagnetic flowmeters, acoustic sensors, and dissolved oxygen sensors. They are reasonably installed at key positions of the sewage pipe network, such as pipe nodes, pump station inlets and outlets, etc., to ensure that the operation status information of the pipe network can be obtained comprehensively and accurately.

[0022] Feature extraction module: This module is based on a multi-scale convolutional neural network to extract spatio-temporal features from the real-time operation data obtained by the data acquisition module. Through a series of complex data processing and feature mining operations, a pipe network state feature matrix is finally generated. This matrix can effectively reflect various feature information during the operation of the pipe network and provide key data support for subsequent anomaly detection.

[0023] Anomaly detection module: Input the pipe network state feature matrix generated by the feature extraction module into a pre-trained spatio-temporal anomaly detection model. This model has been trained with a large amount of historical data and has a powerful pattern recognition ability. It can accurately analyze the pipe network state feature matrix, thereby generating potential leakage area marks and clarifying the areas in the pipe network where leakage may occur.

[0024] Optimization analysis module: According to the leakage area marks given by the anomaly detection module, the optimization analysis module constructs a multi-constraint dynamic optimization model. This model takes pressure balance adjustment and leakage location accuracy as optimization goals and uses an adaptive particle swarm algorithm to globally optimize the pipe network pressure parameters. In this way, not only can the leakage location be determined more accurately, but also the pipe network pressure can be adjusted to reach a balanced state, reducing the leakage risk.

[0025] Hierarchical execution module: According to the optimal regulation strategy output by the dynamic optimization model, the hierarchical execution module realizes the distributed response of instructions through its unique hierarchical execution architecture. This architecture includes a decision-making layer, a regional coordination layer, and an execution layer. The decision-making layer is responsible for generating a global pressure regulation sequence and planning the direction of pipe network pressure regulation from a macroscopic level; the regional coordination layer uses a sliding window optimization algorithm to dynamically match local pressure parameters to make local regulation more accurate and flexible; the execution layer realizes the adjustment of valve opening and pump station power based on the fuzzy proportional integral control algorithm, so as to achieve precise control of the actual operation of the pipe network and ensure the stable operation of the pipe network.

[0026] The present invention will be further described below in conjunction with Embodiment 1 to Embodiment 5: Embodiment 1:

[0027] In this embodiment, the specific implementation manner of the data acquisition module and the related data processing flow are further described in detail. The multi-source sensors used in the data acquisition module include a pressure transmitter, an electromagnetic flowmeter, an acoustic wave sensor, and a dissolved oxygen sensor. During the actual installation process, the pressure transmitter is installed at different positions on the pipeline to measure the pressure values at various points, denoted as ; the electromagnetic flowmeter is installed at an appropriate position to obtain the flow rate data, denoted as . The data acquisition frequencies of these two sensors need to be kept consistent for subsequent processing.

[0028] Perform spatio-temporal alignment on the pressure transmitter data and the electromagnetic flowmeter data to construct a pipeline network pressure-flow distribution map. Spatio-temporal alignment refers to matching the data collected by different sensors at different times and in different spaces so that it can accurately reflect the pressure and flow states at different positions in the pipeline network at the same moment. By dividing time into equal time intervals , and dividing space into certain grid areas , associate the pressure and flow rate data within the same time interval and in the same grid area to construct a pipeline network pressure-flow distribution map.

[0029] For the acoustic wave sensor, the data it collects contains rich pipeline network operation information. Perform wavelet packet decomposition on the acoustic wave sensor data to generate a high-frequency vibration feature sequence. Wavelet packet decomposition is a time-frequency analysis method that can decompose a signal into different frequency bands to extract more refined features. Let the original signal collected by the acoustic wave sensor be , after wavelet packet decomposition, obtain the high-frequency vibration feature sequence , where represents discrete time points, and represents different frequency sub-bands.

[0030] Construct a dual-branch feature fusion network. The first branch uses a three-dimensional convolutional network to extract the spatial correlation features of the pressure-flow distribution map. Divide the pipeline network pressure-flow distribution map into multi-level grids, and each grid cell stores the pressure gradient and the flow rate change rate statistic (i.e., the ratio of the flow rate mutation term to the change time ). In the encoding stage, use a dilated convolutional kernel to expand the spatial perception range. Let the dilation rate of the dilated convolutional kernel be , and the size of the convolutional kernel be , after the dilated convolution operation, the spatial dimension of the feature map changes, thereby obtaining more extensive spatial correlation features. In the decoding stage, the local details are restored through the transposed convolution layer, so that the size and resolution of the feature map are restored to an appropriate state. The second branch uses a temporal convolutional network to extract the time-varying features of the high-frequency vibration feature sequence. Through the convolution operations with different convolution kernel sizes and strides, the variation law of the high-frequency vibration feature sequence over time is mined.

[0031] The spatial correlation features and time-varying features are fused through a cross-modal gating mechanism to generate a joint feature tensor. The cross-modal gating mechanism dynamically adjusts the fusion weights according to the importance of the two features, so that the fused joint feature tensor can better reflect the operation state of the pipe network. Based on the bidirectional long short-term memory network, dynamic pattern modeling is performed on the joint feature tensor. The bidirectional long short-term memory network can consider both past and future information, effectively capture the dynamic patterns in the operation of the pipe network, and finally output a pipe network state feature matrix containing pressure fluctuations, vibration anomalies, and water quality mutations. Example 2:

[0032] This example focuses on the specific construction and operation process of the spatio-temporal anomaly detection model in the anomaly detection module. The spatio-temporal anomaly detection model adopts a multi-level graph neural network structure and generates leakage area markers based on the dynamic graph attention mechanism.

[0033] First, a pipe network-environment interaction graph is constructed. The nodes in the graph include pipe nodes, pump station nodes, valve nodes, and environmental monitoring nodes. Each node has rich attributes. Among them, the attributes of the pipe node include the pressure value , flow value and acoustic wave energy , where represents different pipe nodes. The attributes of the pump station node include pump station power , inlet and outlet pressure difference , etc.; the attributes of the valve node include valve opening , pressure difference before and after the valve , etc.; the attributes of the environmental monitoring node include the surrounding soil humidity , atmospheric pressure , etc.

[0034] The dual-path attention mechanism is adopted. In the first path, the interaction weights between the pipe node and its adjacent nodes are calculated through the spatial graph aggregation layer. Define the relative distance vector between the pipe node and its adjacent nodes, including the hydraulic gradient , pipe material attenuation coefficient and connection direction . The relative distance vector is converted into the offset parameter of the graph convolution kernel through the non-linear mapping layer , and the formula is , where It represents a non - linear mapping function. By superimposing the offset parameter onto the standard graph convolution operation, when calculating the node interaction weights, factors such as the relative positions and connection relationships between nodes can be fully considered. The second path performs feature screening on historical operation data through a temporal graph pooling layer. The temporal graph pooling layer aggregates and screens the historical attribute data of nodes according to the time sequence, extracting key temporal features. For example, through max - pooling or average - pooling operations, features such as the maximum or average pressure value and flow value within a period of time are retained to highlight the changing trend of the data.

[0035] Based on the multi - head graph attention module, the node attributes are iteratively updated, and each attention head fuses the node state and environmental monitoring parameters. Let the number of attention heads be , and each attention head fuses the node attributes and environmental monitoring parameters through different weight matrices. The formula is , where represents the result after fusion by the th attention head, represents the weight of the th attention head for the th parameter. Through skip connections and layer normalization mechanisms, the training process is stabilized. Skip connections can avoid the problem of gradient vanishing, enabling the model to better learn deep features; the layer normalization mechanism normalizes the input of each layer, accelerating the convergence of the model. Finally, after multiple iterations of update and processing, the model outputs the leakage area markers including the pipe network topology constraints, accurately indicating potential leakage areas. Embodiment 3:

[0036] This embodiment focuses on elaborating the specific implementation process of the adaptive particle swarm optimization algorithm in the optimization analysis module. In the optimization analysis module, the pressure optimization problem is modeled as a multi - constraint non - linear programming problem. The decision variables include discrete valve state variables and continuous pump station power variables . The discrete valve state variable represents the on - off state of the valve. For example, represents the valve being closed, represents the valve being open; the continuous pump station power variable can vary continuously within a certain range to meet different pipe network pressure requirements.

[0037] Initialize the particle swarm and calculate the Pareto - front candidate solutions. The particle swarm consists of multiple particles, and each particle represents a set of possible decision variable values, that is, a combination of valve states and pump station powers. Let the number of particles be , and the position of the th particle is represented as Calculate the fitness value of each particle. The fitness value is used to measure the degree to which the combination of decision variables represented by the particle meets the optimization goal. Here, the optimization goals are pressure balance adjustment and leakage location accuracy. An inertia weight adaptive mechanism is adopted to adjust the particle search range according to the real-time pipe network load. Collect the particle fitness values and constraint violation degrees as training data to construct an extreme learning machine network to fit the weight coefficients and the mapping relationship with the pipe network state (such as parameters like pressure and flow rate). The formula is , where represents the fitting function of the extreme learning machine network. Update the network parameters through the online recursive least squares method to adjust the particle swarm search step in real time. When a sudden change in the pipe network pressure is detected, trigger the weight rapid reset operation so that the particle swarm can quickly adapt to the change in the pipe network state and search for a better solution again.

[0038] In the optimization stage, based on the objective decomposition technology, divide the global problem into multiple sub-group optimization problems. Divide the particle swarm into several sub-groups, and each sub-group focuses on optimizing one sub-objective or one local area of the optimization problem. For example, one sub-group can be mainly responsible for optimizing the pressure balance in a certain area, and another sub-group focuses on improving the leakage location accuracy. In the constraint repair stage, introduce a relaxation factor to balance the conflicting constraints between sub-groups. When the optimization goals of different sub-groups conflict, adjust the relaxation factor so that the optimization results of each sub-group can be coordinated with each other, avoiding the situation where over-optimizing a certain sub-objective leads to the inability to achieve the overall goal. Adopt a parallel evolution strategy to iteratively optimize the sub-groups, update the local optimal solution each time and synchronize the optimal solution through global information sharing. Each sub-group performs iterative optimization within its own search space to find the local optimal solution, and then through the global information sharing mechanism, communicate and integrate the optimal solutions of each sub-group, enabling the entire particle swarm to evolve in a better direction and finally achieving the global optimization of the pipe network pressure parameters. Example 4:

[0039] This example details the specific implementation of the regional coordination layer in the hierarchical execution module using the sliding window optimization algorithm to dynamically match the local pressure parameters. When constructing the regional coordination layer, a time-varying pipe network pressure propagation model is built, and the hydrodynamic equation is discretized into a state transition equation. Let the pressure at a certain point in the pipeline at time be , the flow rate be , the pressure decay gradient be , the flow rate mutation term be , and the valve response delay term be , then the state transition equation is , where represents the spatial step size, represents the time step size. This equation describes the variation law of pressure in the pipe network with time and space, considering factors such as pressure attenuation, flow rate change, and valve response delay.

[0040] The design window optimization objective function includes a pressure fluctuation suppression term, an energy consumption minimization term, and a leakage location error correction term. Let the pressure fluctuation suppression term be , where represents the pressure fluctuation value at the -th time step, represents the number of time steps within the sliding window. This term is used to measure and suppress the degree of pressure fluctuation, making the pipe network pressure more stable. The energy consumption minimization term is , where represents the power of the pumping station at the -th time step, represents the number of time steps during which the pumping station operates within the sliding window. This term is used to calculate the total energy consumption of the pumping station within the sliding window. By optimizing this term, the energy consumption cost of the pipe network operation can be reduced. The leakage location error correction term is , where represents the actual leakage location, represents the predicted leakage location, represents the number of leakage points. This term is used to measure the accuracy of leakage location. By minimizing this term, the accuracy of leakage location can be improved. Combining these three terms, the window optimization objective function is , where , , are the weight coefficients of the three terms respectively, which are adjusted according to actual needs to balance the importance of different objectives. By continuously adjusting the local pressure parameters within the sliding window, the value of the objective function is minimized, thereby achieving dynamic matching of local pressure parameters and improving the stability and efficiency of the pipe network operation. Example 5:

[0041] This example details the specific process of the execution layer in the hierarchical execution module to implement the adjustment of valve opening and pumping station power based on the fuzzy proportional integral control algorithm. The execution layer designs a multi-level fuzzy inference system to map the pressure error and the flow rate error to the fuzzy rule triggering degree. The pressure error refers to the difference between the actual pressure value and the set pressure value, and the flow rate error It refers to the difference between the actual flow value and the set flow value. First, the pressure error and the flow error are fuzzified and divided into different fuzzy sets. For example, the pressure error is divided into fuzzy sets such as "negative large", "negative medium", "negative small", "zero", "positive small", "positive medium", "positive large", etc., and the flow error is divided similarly. Fuzzy rules are formulated based on experience and actual operation data. For example, when the pressure error is "positive large" and the flow error is "positive small", the corresponding fuzzy rule is triggered to determine the adjustment direction and degree of the valve opening and the pump station power.

[0042] An adaptive proportional-integral parameter adjustment mechanism is constructed to dynamically adjust the slope of the fuzzy membership function according to the error accumulation. Let the error accumulation be , where can be the pressure error or the flow error, and represents the number of time steps. When the error accumulation is large, it indicates that the deviation of the system is large and the adjustment speed needs to be increased. At this time, the slope of the fuzzy membership function is increased, making the fuzzy inference system more sensitive to the error and able to make adjustment responses more quickly; when the error accumulation is small, the slope of the fuzzy membership function is appropriately reduced to make the adjustment process smoother and avoid over-adjustment. By continuously monitoring the error accumulation and dynamically adjusting the slope of the fuzzy membership function according to its change, the precise adjustment of the valve opening and the pump station power is achieved. For example, let the fuzzy membership function be , where is the input variable (pressure error or flow error), is the slope, and according to the size of the error accumulation , the value of is adjusted through a certain functional relationship, such as , where

[0043] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0044] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A leakage detection system for urban sewage pipe networks, characterized in that, It includes: Data acquisition module: used to collect real-time operation data of the pipe network through multi-source sensors, including pressure, flow rate, acoustic signal, and water quality parameters; Feature extraction module: based on a multi-scale convolutional neural network, extract spatio-temporal features from the real-time operation data to generate a pipe network state feature matrix; Anomaly detection module: input the pipe network state feature matrix into a pre-trained spatio-temporal anomaly detection model to generate potential leakage area marks; Optimization analysis module: construct a multi-constraint dynamic optimization model according to the leakage area marks, with the optimization objectives of pressure balance adjustment and leakage location accuracy, and use an adaptive particle swarm optimization algorithm to globally optimize the pipe network pressure parameters; Hierarchical execution module: based on the dynamic optimization model, output the optimal control strategy, and achieve distributed response of instructions through a hierarchical execution architecture, which includes a decision-making layer, a regional coordination layer, and an execution layer. Among them, the decision-making layer generates a global pressure control sequence, the regional coordination layer uses a sliding window optimization algorithm to dynamically match local pressure parameters, and the execution layer adjusts the valve opening and pump station power based on a fuzzy proportional integral control algorithm.

2. The urban sewage pipe network leakage detection system according to claim 1, characterized in that The collection of real-time operation data of the pipe network through multi-source sensors includes: The multi-source sensors include a pressure transmitter, an electromagnetic flowmeter, an acoustic wave sensor, and a dissolved oxygen sensor; Perform spatio-temporal alignment on the pressure transmitter data and the electromagnetic flowmeter data to construct a pipe network pressure-flow distribution map; perform wavelet packet decomposition on the acoustic wave sensor data to generate a high-frequency vibration feature sequence; Construct a dual-branch feature fusion network. The first branch uses a three-dimensional convolutional network to extract the spatial correlation features of the pressure-flow distribution map, and the second branch uses a temporal convolutional network to extract the time-varying features of the high-frequency vibration feature sequence; Fuse the spatial correlation features and time-varying features through a cross-modal gating mechanism to generate a joint feature tensor; perform dynamic mode modeling on the joint feature tensor based on a bidirectional long short-term memory network, and output a pipe network state feature matrix including pressure fluctuations, vibration anomalies, and water quality mutations.

3. The urban sewage pipe network leakage detection system according to claim 1, characterized in that, The spatio-temporal anomaly detection model adopts a multi-level graph neural network structure and generates leakage area marks based on a dynamic graph attention mechanism; The multi-level graph neural network structure includes: Construct a pipe network-environment interaction graph. The nodes in the graph include pipe nodes, pump station nodes, valve nodes, and environmental monitoring nodes, and the node attributes include pressure values, flow rate values, and acoustic wave energy; Adopt a dual-path attention mechanism. The first path calculates the interaction weights between pipe nodes and adjacent nodes through a spatial graph aggregation layer, and the second path performs feature screening on historical operation data through a temporal graph pooling layer; Iteratively update the node attributes based on a multi-head graph attention module. Each attention head fuses the node state and environmental monitoring parameters; stabilize the training process through a skip connection and a layer normalization mechanism, and finally output a leakage area mark including pipe network topology constraints.

4. The urban sewage pipe network leakage detection system according to claim 1, characterized in that The adaptive particle swarm optimization algorithm integrates an inertia weight adjustment and a constraint repair strategy, including: Model the pressure optimization problem as a multi-constraint non-linear programming problem, and the decision variables include discrete valve state variables and continuous pump station power variables; Initialize the particle swarm and calculate the Pareto front candidate solutions, and adopt an inertia weight adaptive mechanism to adjust the particle search range according to the real-time pipe network load; In the optimization stage, divide the global problem into multiple sub-group optimization problems based on the objective decomposition technology; in the constraint repair stage, introduce a relaxation factor to balance the conflicting constraints between sub-groups; Adopt a parallel evolution strategy to iteratively optimize the sub-groups, update the local optimal solutions in each iteration, and synchronize the optimal solutions through global information sharing.

5. The urban sewage pipe network leakage detection system according to claim 1, characterized in that, The regional coordination layer adopts a sliding window optimization algorithm to dynamically match the local pressure parameters, including: Construct a time-varying pipe network pressure propagation model, discretize the hydrodynamic equation into a state transition equation, and the equation includes a pressure attenuation gradient, a flow rate mutation term, and a valve response delay term; Design a window optimization objective function, including a pressure fluctuation suppression term, an energy consumption minimization term, and a leakage location error correction term.

6. The urban sewage pipe network leakage detection system according to claim 1, characterized in that The execution layer realizes the regulation of the valve opening and the pump station power based on the fuzzy proportional integral control algorithm, including: Design a multi-level fuzzy inference system to map the pressure error and the flow rate error to the fuzzy rule triggering degree; construct an adaptive proportional integral parameter adjustment mechanism to dynamically adjust the slope of the fuzzy membership function according to the error accumulation amount.

7. The urban sewage pipe network leakage detection system according to claim 2, characterized in that The three-dimensional convolutional network adopts a multi-scale dilated convolutional structure to accelerate feature extraction, including: Divide the pipe network pressure-flow distribution map into multi-level grids, and each grid cell stores the pressure gradient and the flow rate change rate statistics; In the encoding stage, use dilated convolutional kernels to expand the spatial perception range, and in the decoding stage, restore the local details through transposed convolutional layers; Introduce a spatial attention module to assign regional weights to the feature maps.

8. The urban sewage pipe network leakage detection system according to claim 3, characterized in that The spatial graph aggregation layer adopts a relative distance encoding mechanism, including: Define the relative distance vector between the pipeline node and the adjacent nodes, including the hydraulic gradient, the pipeline material attenuation coefficient, and the connection direction; Convert the relative distance vector into the offset parameter of the graph convolutional kernel through a non-linear mapping layer; Superimpose the offset parameter on the standard graph convolutional operation.

9. The urban sewage pipe network leakage detection system according to claim 4, characterized in that, The inertia weight adaptive mechanism is realized based on an online feedback strategy, including: Collect the particle fitness values and constraint violation degrees in the historical optimization process as training data; Construct an extreme learning machine network to fit the mapping relationship between the weight coefficients and the pipe network state; Update the network parameters through the online recursive least squares method, and adjust the particle swarm search step size in real time; When a sudden change in the pipe network pressure is detected, trigger a rapid weight reset operation.

10. A method for detecting leakage in an urban sewage pipe network, applied to the urban sewage pipe network leakage detection system according to any one of claims 1 to 9, characterized in that, It includes the following steps: Step 1: Use a multi-source sensor to collect the real-time operation data of the pipe network through a data acquisition module, and the real-time operation data includes pressure, flow rate, acoustic wave signal, and water quality parameters; Step 2: Use a feature extraction module to perform spatio-temporal feature extraction on the real-time operation data collected in Step 1 based on a multi-scale convolutional neural network to generate a pipe network state feature matrix; Step 3: Input the pipe network state feature matrix generated in Step 2 into a pre-trained spatio-temporal anomaly detection model, and generate potential leakage area marks through an anomaly detection module; Step 4: The optimization analysis module constructs a multi-constraint dynamic optimization model based on the leakage area markers obtained in Step 3. The model takes pressure balance adjustment and leakage location accuracy as optimization objectives, and uses an adaptive particle swarm optimization algorithm to globally optimize the pipe network pressure parameters; Step 5: Based on the optimal regulation strategy output by the dynamic optimization model in Step 4, the hierarchical execution module of the hierarchical execution architecture realizes the distributed response of instructions. Among them, the decision-making layer generates a global pressure regulation sequence, the regional coordination layer uses a sliding window optimization algorithm to dynamically match local pressure parameters, and the execution layer realizes the adjustment of valve opening and pump station power based on the fuzzy proportional integral control algorithm.

Images