Modeling analysis method for optimizing urban water supply pipe network

Abstract

The invention discloses a modeling analysis method for optimizing an urban water supply pipe network, and belongs to the technical field of intelligent water affairs, and the method comprises the steps: building a comprehensive pipe network database, collecting hydraulic parameters through Internet of Things equipment, and fusing the hydraulic parameters with GIS topological data; constructing a dynamic hydraulic simulation model, coupling real-time data of a pressure sensor with an EPANET model, and correcting a friction resistance coefficient of a pipe section; a hybrid optimization algorithm architecture is designed, a double-target genetic algorithm is adopted for old pipe network transformation, and a constrained particle swarm algorithm is adopted for new pipe network planning; executing multi-stage optimization calculation, generating a pipe diameter feasible solution space in combination with geographical constraints, and iteratively outputting a scheme meeting a pressure threshold value; and generating a pipe network optimal configuration scheme, and outputting an engineering implementation list through spatial overlay analysis and valve regulation logic. According to the method, the problem of modeling distortion caused by a multi-source data island is solved, the defect that a static model is difficult to adapt to dynamic water demands is overcome, and the economical efficiency and the engineering implementability of a pipe network optimization scheme are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent water services, and particularly to a modeling and analysis method for optimizing urban water supply pipe networks. Background Art

[0002] As a key infrastructure, the operation efficiency of urban water supply pipe networks directly affects people's livelihood guarantee and resource sustainability. In the renovation of old urban pipe networks, the leakage rate caused by pipe aging is generally high, and it is urgent to accurately locate high-loss pipe sections and optimize replacement strategies; in the planning of new urban areas, a balance needs to be achieved between construction costs and water supply reliability. Existing technologies mainly rely on manual experience to design renovation plans and are difficult to meet the complex optimization requirements of large-scale pipe networks.

[0003] There are many problems in existing technologies. There are defects in data integration. Traditional modeling relies on static GIS data and manual inspection records, and there is a lack of an effective fusion mechanism for Internet of Things real-time data (pressure, flow) and spatial data. In urban intelligent water service projects, due to the differences in the SCADA system and GIS coordinate system, many monitoring points cannot be matched with topological nodes, resulting in distorted hydraulic model parameters. Existing models have insufficient adaptability. The mainstream hydraulic models adopt fixed friction coefficients and do not consider the dynamic changes caused by pipe aging. The optimization efficiency of the model for pipe networks is low. In the renovation of old pipe networks, discrete decision variables (such as pipe diameter combinations) need to be processed, and the traditional enumeration method takes more than 72 hours to calculate when the scale is above 1000 pipe sections.

[0004] Existing technologies have many limitations. At the data level, the spatio-temporal benchmarks of multi-source heterogeneous data are not unified, and the utilization rate of real-time monitoring data is insufficient; at the model level, the static parameter setting cannot reflect dynamic working conditions such as valve regulation and water consumption fluctuations. At the collaborative optimization level, single-objective optimization algorithms are difficult to coordinate economic (renovation cost) and reliability (pressure stability) goals, and the feasibility of the solution set is poor. These defects make it difficult for existing methods to support high-precision and high-efficiency pipe network optimization decisions, and there is an urgent need to develop a technical solution that collaborates the entire chain of data-model-optimization. Summary of the Invention

[0005] The present invention overcomes the problem of modeling distortion caused by multi-source data islands in existing technologies, overcomes the defect that static models are difficult to adapt to dynamic water consumption demands, and significantly improves the economy and engineering feasibility of pipe network optimization solutions.

[0006] To achieve the above object, the present invention adopts the following scheme: A modeling and analysis method for optimizing urban water supply pipe networks, which includes the following steps: S1. Obtain multi-source heterogeneous data, including: collecting hydraulic parameters of pressure sensors and flow meters in the target pipe network through Internet of Things devices, and synchronously obtaining pipe network topology data, pipe material attribute data, and historical user water consumption record data in the geographic information system; registering the obtained multi-source heterogeneous data according to a unified spatio-temporal coordinate system and storing it in a relational database to establish a comprehensive pipe network database; S2. Based on the pipe network topology data, establish a node-segment topology relationship matrix, couple the real-time monitoring data of pressure sensors with the EPANET hydraulic calculation model, correct the pipe segment friction coefficient through the non-linear least squares method, generate a three-dimensional pipe network hydraulic model including pipe diameter, elevation, and valve status, and construct a dynamic hydraulic simulation model; S3. For the old pipe network renovation requirements, establish a fitness function with the pipe network renovation cost and the reduction of leakage water volume as dual objectives, and use a genetic algorithm with an elite retention strategy to sort the priority of pipe segment replacement; for the new pipe network planning requirements, establish a particle swarm optimization algorithm with construction cost and hydraulic stability as constraint conditions to obtain a hybrid optimization algorithm framework; S4. Perform multi-stage optimization calculations, input the pipe material attribute data in the comprehensive pipe network database into the dynamic hydraulic simulation model for initial hydraulic balance calculations, generate a feasible solution space of pipe diameters according to the geographical constraint conditions in the old pipe network renovation area, and iteratively calculate through the hybrid optimization algorithm framework to obtain a pipe diameter combination plan that meets the node pressure threshold; S5. Conduct a spatial overlay analysis of the optimized pipe diameter combination plan and the road alignment data in the geographic information system, and combine the valve position data to generate an engineering implementation list including pipe segment replacement coordinates, new pipeline paths, and valve control instructions, and generate a pipe network optimization configuration plan.

[0007] Preferably, the establishment of the comprehensive pipe network database in step S1 specifically includes the following steps: Install Internet of Things pressure sensors at the intersections of pipes in the target pipe network, install ultrasonic flow meters at the user water inlet, and transmit the collected hydraulic parameters to the edge computing gateway in real time through the LoRaWAN communication protocol; Convert the pipe network topology data stored in the geographic information system into a vector layer in the WGS84 coordinate system, perform geographic coding on the address information in the historical user water consumption record data, and establish a spatial mapping relationship with the pipe network topology nodes; For the real-time hydraulic parameter data stream collected by Internet of Things devices, use the cubic spline interpolation algorithm to reconstruct the data segment with missing timestamps, and conduct a spatial overlay analysis with the pipe network topology data in the geographic information system to generate a three-dimensional spatio-temporal data cube including pipe network node spatial coordinates, pipe segment elevation data, and corresponding time series hydraulic parameters; Establish a calculation model for the theoretical value of pipeline pressure. When the deviation between the measured value of the sensor and the theoretical value exceeds the safety threshold corresponding to the pipeline pressure-bearing grade, automatically trigger the marking of abnormal data, and use the quartile method to clean the abnormal data, retaining the valid data within the upper and lower quartile intervals; Store the registered pipe network topology data in the PostGIS spatial database, and store the time series hydraulic parameter data in the time series database. Establish an association index for the two databases by creating a unique identification code for the pipe network nodes, where the unique identification code for the pipe network nodes is generated by combining the pipe network number, administrative division code, and spatial coordinate hash value; Set the weekly incremental update cycle for geographic information system data, perform a moving average filtering process with a 5-minute time window on the real-time data stream of the Internet of Things, and automatically trigger the re-registration calculation of the pipe network topology data and hydraulic parameters when the user water consumption record data is updated.

[0008] Preferably, the following method is used to generate the pipe network optimization configuration plan in step S5: Convert the pipe section replacement coordinates in the optimized pipe diameter combination plan into the Shapefile format of the geographic information system, perform a buffer analysis with the road centerline vector layer, set 1.2 meters outside the road red line as the construction safety distance, and eliminate the pipe line paths that overlap with the existing underground pipe gallery by more than 30%; establish a node-valve association matrix based on the valve position data in the pipe network topology data, and use the breadth-first search algorithm to determine the control influence range of each valve, generating a control instruction tree diagram including the valve opening and closing sequence and time series; Extract the pipe material attribute data from the comprehensive pipe network database, combine the lifting machinery operation radius parameters at the construction site, perform a curvature radius check on the pipe section paths with a pipe diameter ≥ 600mm, and insert a transition elbow node when the turning angle exceeds the minimum turning radius of the excavator; associate the new pipe line path after the spatial overlay analysis with the construction feasibility check result, sort it according to the unique identification code of the pipe network node, and output a structured table including the pipe section number, start coordinate, end coordinate, pipe diameter specification, valve control time series, and construction machinery model; When the road alignment data changes, trigger the recalculation of the spatial overlay analysis, and automatically update the coordinate data of the affected pipe sections by comparing the version numbers; use the constraint satisfaction algorithm to check the spatio-temporal conflicts between the pipe section replacement and new construction paths in the project implementation list. When two construction tasks are detected at the same coordinate point, automatically postpone the execution time of the non-urgent task and generate a conflict warning report; merge and encode the project implementation list and the valve control instruction tree diagram to generate an XML format document and ensure data integrity through digital signature.

[0009] Preferably, the following method is used to construct a dynamic hydraulic simulation model in step S2: Extract the coordinates of the pipe segment endpoints based on the pipe network topology data of the geographic information system, construct a 0-1 incidence matrix with the pipe network nodes as row vectors and the adjacent pipe segments as column vectors, where the flow direction of the pipe segment is represented by the positive and negative values determined by the flowmeter data; Divide the pressure data monitored in real time by the pressure sensors into data slices at a 15-minute time granularity, inject them into the hydraulic calculation model through the Toolkit API interface of EPANET, and establish a dynamic mapping relationship between the measured pressure values and the calculated values; Aiming at the problem of the initial value setting error of the pipe segment friction coefficient, use the Levenberg-Marquardt algorithm as the iterative optimizer of the nonlinear least squares method, take the root mean square error between the measured pressure data slices and the EPANET calculated values as the objective function, and synchronously optimize the friction coefficient combinations of three adjacent pipe segments to complete the friction coefficient correction; Substitute the corrected friction coefficient into the EPANET model for transient simulation, extract the flow velocity and head loss data of each pipe segment at different valve openings, construct a pipe network hydraulic state tensor with spatial coordinates as the X-Y axis and time series as the Z axis, and generate a three-dimensional hydraulic characteristic field; Close the pressure regulating valve of the target pipe network during the low water consumption period every day, obtain the actual pressure distribution data through the SCADA system, and perform the Kolmogorov-Smirnov test on the simulation results of the three-dimensional hydraulic characteristic field. When the P value is less than 0.05, re-correct the friction coefficient.

[0010] Preferably, the process of collaborative correction of the friction coefficient includes: Construct a dynamic water use pattern library, use a long short-term memory neural network to train the historical record data of user water consumption, generate a water consumption prediction model including weekdays, holidays and seasonal changes, and output the water consumption change curves of each node in the next 24 hours; Discretize the water consumption prediction curve into the boundary conditions of the node water demand at 15-minute intervals, and introduce a dynamic weight factor in the iterative process of the Levenberg-Marquardt algorithm to make the optimization process preferentially match the measured pressure data slices during the peak water consumption period; Establish a spatial constraint of the friction coefficient. Determine the change range of the friction coefficient of pipe segments with different materials according to the pipe material attribute data. When the optimized friction coefficient exceeds the expected friction coefficient range of the pipe material, trigger an abnormal state warning for the pipe segment; Set a double-threshold convergence criterion, and terminate the optimization when the change amount of the root mean square error in three consecutive iterations is less than 0.01 m and the change amount of the friction coefficient is less than 0.0001, and retain the optimal friction coefficient combination; Write the finally corrected friction coefficient into the basic input file of the EPANET model, update the pipe material attribute data in the comprehensive pipe network database at the same time, and record the version number and check code.

[0011] Preferably, the execution of multi-stage optimization calculation in step S4 includes the following steps: Extract the road width data, the distribution data of the underground comprehensive pipe gallery, and the coordinate data of existing buildings in the old pipe network renovation area, establish a two-dimensional rasterized constraint matrix including pipe section burial depth limits, horizontal offset tolerances, and construction restricted area ranges, and obtain the geographical constraint condition matrix; according to the pipe material attribute data in the comprehensive pipe network database, screen out the set of pipe diameter specifications that meet the construction standards, perform a Cartesian product operation in combination with the geographical constraint condition matrix, and exclude the pipe diameter-path combinations with a distance less than 1.5 meters from the underground power pipeline to construct a feasible solution space for pipe diameters; Implement parallel computing of the hybrid algorithm, embed the velocity update formula of the particle swarm optimization algorithm into the genetic algorithm with an elite retention strategy, divide the feasible solution space of pipe diameters into 4 sub-spaces, calculate the initial fitness values of each sub-space respectively, and retain the solution set with the top 10% of the fitness values; Execute the pressure threshold verification, compare the node pressure simulation values output by the dynamic hydraulic simulation model with the minimum service pressure values in the urban water supply specification node by node. When there are more than 5% of the nodes that do not meet the standards in three consecutive iteration cycles, expand the feasible solution space of pipe diameters; arrange the pipe diameter combination schemes that pass the pressure threshold verification in ascending order of the renovation cost, output the top three candidate schemes and their corresponding hydraulic stability indicators to obtain the optimized scheme set.

[0012] Preferably, constructing the feasible solution space of pipe diameters includes the following steps: Establish a multi-objective conflict resolution model, adopt the Pareto front analysis method to handle the trade-off relationship between the renovation cost and hydraulic stability in pipe diameter selection, and set the condition that the increase rate of the renovation cost does not exceed 1.5 times the improvement rate of hydraulic stability as the effective solution screening condition; use historical project data to train a random forest classifier to predict the construction feasibility probability of each pipe diameter-path combination, and automatically exclude the combination when the predicted value is lower than 0.7; Generate a dynamic solution space index, establish a B+ tree index structure for the feasible solution space of pipe diameters according to the administrative division code, and combine the unique identification code of the pipe network node to retrieve and update the solution space data; adopt the principal component analysis method to extract the key feature vectors of the pipe diameter combination scheme, reduce the solution space dimension from the original number of pipe sections n to the log2(n) level; randomly select 5% of the feasible solutions for on-site inspection verification, and trigger the regeneration of the geographical constraint condition matrix when the non-pass rate of the verification exceeds 20%.

[0013] Preferably, implementing parallel computing of the hybrid algorithm includes the following specific steps: Construct a distributed computing architecture, use the MPI parallel computing framework to allocate 4 sub-spaces to different computing nodes, each node independently runs the hybrid optimization algorithm and synchronizes the optimal solution data every 20 minutes; dynamically adjust the crossover probability of the genetic algorithm according to the dispersion degree of the sub-space solution set, and reduce the crossover probability from 0.8 to 0.6 when the standard deviation of the solution set exceeds the set threshold; Monitor the change rate of the fitness values of each subspace. When the change rate is less than 0.1% within three consecutive calculation cycles, terminate the iterative calculation of this subspace in advance; perform non-dominated sorting on the optimal solutions of the four subspaces, and use the crowding degree comparison operator to select evenly distributed Pareto optimal solutions to form the final optimization scheme set; when the solution set of a certain subspace is empty due to geographical constraint conflicts, relax the pipe burial depth limit of this area by 10% and restart the calculation process.

[0014] Preferably, the construction of the dynamic hydraulic simulation model in step S2 includes the following specific steps: Establish a dynamic data fusion interface, convert the pipe network topology data of the geographic information system into the GraphML format with topology verification, where the pipe segment connection relationship is stored through an adjacency list, and the node elevation data is supplemented by digital elevation model interpolation; implement real-time data coupling, perform Kalman filtering on the original pressure data stream collected by the pressure sensor to eliminate the transient fluctuations caused by the water hammer effect, generate the smoothed pressure time series data, and inject the hydraulic model boundary conditions through the INP file interface of EPANET; Execute multi-parameter joint calibration, use the constrained nonlinear least squares method, take the pipe segment friction coefficient and the basic water demand of the node as the synchronous optimization variables, and set the physical constraint interval of the initial value of the pipe friction coefficient; Construct a three-dimensional hydraulic response surface, simulate the step change process of the valve opening from 10% to 90% through the transient analysis module of EPANET, record the gradient matrix of the flow velocity of each pipe segment with the valve state change, and generate a three-dimensional hydraulic characteristic space composed of the pipe diameter, elevation and valve opening; During the stable water use period in the late night every day, close 50% of the regulating valves in the target area, collect the actual pressure distribution data through the SCADA system, and perform dynamic time warping matching with the simulated values in the three-dimensional hydraulic characteristic space. When the mean absolute error exceeds 0.05 MPa, trigger the iterative correction of the friction coefficient.

[0015] Preferably, the execution of multi-parameter joint calibration includes the following steps: Construct a dynamic decomposition model of the water use pattern, use the variational mode decomposition algorithm to decompose the historical record data of the user water consumption into 6 intrinsic mode function components, and extract the time-frequency characteristics of the daily cycle component and the random fluctuation component; Introduce a time-varying weight factor into the objective function of the nonlinear least squares method to increase the weight of the measured data in the corresponding period of the daily cycle component to 2.5 times that of the random fluctuation component; according to the service life data in the pipe material attribute database, set the upper limit of the friction coefficient optimization to 0.022 for cast iron pipe sections over 20 years old, and set the lower limit of the optimization to 0.009 for polyethylene pipe sections replaced within 5 years; use the MPI parallel computing framework to divide the pipe network into 4 hydraulic zones, and each zone independently runs the Levenberg-Marquardt optimizer to achieve global convergence through the exchange of boundary node pressure data; generate a parameter credibility report, record the correlation coefficient between the change in the friction coefficient and the pressure error reduction rate in each iteration, and mark the pipe section as a potential abnormal pipe section and output the spatial coordinates when the absolute value of the correlation coefficient is less than 0.7.

[0016] The present invention has at least the following beneficial effects: (1) Through the multi-source spatio-temporal data fusion mechanism, the data island problem is solved, the integrity of the modeling data is significantly improved, the dynamic hydraulic model couples the measured data and parameter correction, and the calibration error of the friction coefficient is reduced; (2) The combination of the data quality verification and model verification mechanisms eliminates the interference of abnormal data on the optimization results and improves the stability of the scheme; (3) The spatial overlay analysis and construction parameter integration reduce the scheme change rate in the project implementation stage; (4) The dynamic weight factor and the parallel optimization framework cooperate to greatly compress the calculation time of large-scale pipe networks; (5) The solution space compression and conflict adaptive repair improve the convergence speed of the optimization algorithm on the premise of ensuring the quality of the scheme. Brief Description of the Drawings

[0017] Figure 1 It is a principle flow chart of the modeling and analysis method for urban water supply network optimization provided by the present invention. Detailed Embodiments

[0018] The following further describes the present invention in detail with reference to the drawings so that those skilled in the art can implement it according to the description in the specification.

[0019] As Figure 1 shown, the modeling and analysis method for urban water supply network optimization provided by the present invention includes the following steps: S1. Obtain multi-source heterogeneous data, including: collect the hydraulic parameters of pressure sensors and flow meters in the target pipe network through Internet of Things devices, and synchronously obtain the pipe network topology data, pipe material attribute data, and historical user water consumption record data in the geographic information system; register the obtained multi-source heterogeneous data according to the unified spatio-temporal coordinate system and store it in the relational database to establish a comprehensive pipe network database.

[0020] By deploying Internet of Things pressure sensors (such as piezoresistive sensors) at key nodes of the target pipeline network (such as pipeline intersections, pump station outlets) and ultrasonic flowmeters at the user's water inlet, the pipeline pressure value (unit: MPa) and the instantaneous flow value (unit: m 3 / h) are collected in real time at a sampling frequency of 1 - 5 minutes. Synchronously extract vector-format pipeline network topology data (including the start / end coordinates of pipeline segments, connection relationships), pipe material attribute data (such as pipeline material type, inner diameter size, roughness coefficient, service life), and user water consumption historical record data (water consumption sequences counted by day / hour) from the geographic information system. Perform spatio-temporal benchmark unification processing on the above multi-source heterogeneous data: convert the GIS topology data into point-line layers in the WGS84 coordinate system; perform geocoding on the text addresses in the water consumption records; associate the Internet of Things real-time data stream with the spatial data through timestamp alignment. Finally, store the registered data in a relational database (such as PostgreSQL) to form a comprehensive data table containing spatial coordinate fields, timestamp fields, hydraulic parameter fields, and pipe material attribute fields.

[0021] S2. Establish a node-segment topology relationship matrix based on the pipeline network topology data, couple the real-time monitoring data of the pressure sensor with the EPANET hydraulic calculation model, correct the pipe segment friction coefficient through the nonlinear least squares method, generate a three-dimensional pipeline network hydraulic model including pipe diameter, elevation, and valve status, and construct a dynamic hydraulic simulation model.

[0022] Establish a node-segment topology relationship matrix based on the pipeline network topology data. This matrix is in the form of a sparse matrix, where the row index corresponds to the pipeline network node number, the column index corresponds to the pipe segment number, and the matrix element value of 1 indicates the connection between the node and the pipe segment, and -1 indicates the opposite flow direction. Through the programming interface of the EPANET hydraulic calculation engine (such as Toolkit API), dynamically inject the pressure data stream (sliced at a 15-minute time granularity) monitored by the pressure sensor into the model boundary conditions. For the problem of setting errors in the pipe segment friction coefficient in the traditional model (such as the C value in the Hazen-Williams formula), the nonlinear least squares method is used for collaborative correction: taking the minimization of the root mean square error between the measured pressure slice and the EPANET calculated value as the optimization goal, and using the Levenberg-Marquardt algorithm to iteratively adjust the friction coefficient combination of adjacent pipe segments (optimizing 3 - 5 associated pipe segments each time). Substitute the corrected parameters into the EPANET transient simulation module to output a hydraulic model with three-dimensional characteristics: spatial dimension (X-Y axis is the node coordinate), attribute dimension (Z-axis elevation and pipe diameter value), and dynamic dimension (flow velocity / pressure distribution under different valve openings).

[0023] S3. For the requirement of old pipe network renovation, establish a fitness function with the renovation cost of the pipe network and the reduction in leakage water volume as the dual objectives, and use a genetic algorithm with an elite retention strategy to sort the priority of pipe section replacement; for the requirement of new pipe network planning, establish a particle swarm optimization algorithm with construction cost and hydraulic stability as the constraint conditions to obtain a hybrid optimization algorithm framework.

[0024] For the old pipe network renovation scenario, construct a dual-objective fitness function: Objective 1 is the pipe section replacement cost (including pipe material cost and construction cost), and Objective 2 is the reduction in leakage water volume (the leakage simulation value calculated based on the hydraulic model). Use a genetic algorithm with an elite retention strategy for optimization: Initialize the population as the combination of pipe sections to be renovated, adopt two-point crossover for the crossover operation, and randomly replace the pipe diameter specifications with a probability of 0.1 - 0.3 for the mutation operation. Retain the top 10% of the elite solutions in terms of fitness in each generation. For the new pipe network planning scenario, establish a particle swarm optimization algorithm with construction cost (total pipe material price + construction cost) and hydraulic stability (variance of node pressure fluctuation) as the constraint conditions: The particle position vector represents the pipe diameter combination, and the velocity update formula introduces an inertia weight factor (with a value range of 0.4 - 0.9). The calculation results of the dynamic hydraulic model are shared by the two types of algorithms, and the high-priority renovation pipe sections output by the optimization of the old pipe network (such as pipe sections with a leakage risk > 5%) will be used as the constraint conditions for the initial solution space of the new pipe network planning.

[0025] S4. Perform multi-stage optimization calculations, input the pipe material attribute data in the comprehensive pipe network database into the dynamic hydraulic simulation model for initial hydraulic balance calculations, generate a feasible solution space for pipe diameters according to the geographical constraint conditions of the old pipe network renovation area, and obtain the pipe diameter combination plan that meets the node pressure threshold through iterative calculations using the hybrid optimization algorithm framework.

[0026] First, input the pipe material property data (such as pipe diameter and friction coefficient) in the integrated pipe network database into the dynamic hydraulic model and perform the initial hydraulic balance calculation (simulating the flow / pressure distribution under the unmodified state). Generate a feasible solution space for pipe diameters according to the geographical constraint conditions in the old pipe network renovation area: Extract road width data (requirement: ≥4 meters for construction), the location of the underground integrated pipe gallery (horizontal distance from the power pipeline ≥1.5 meters), and building coordinates (construction restricted area radius ≥3 meters) to form a two-dimensional rasterized constraint matrix; Screen the pipe diameter specifications that meet the GB / T 50268 standard (such as ductile iron pipes with DN100 - DN600), and exclude the pipe diameter - path combinations that violate the geographical constraints through Cartesian product operation. Perform iterative calculation through the hybrid optimization algorithm framework: The genetic algorithm performs selection - crossover - mutation operations within the feasible solution space, and the particle swarm algorithm optimizes the solution set of the subspace (divide into 4 - 6 sub-regions); In each round of iteration, call the dynamic hydraulic model to verify whether the node pressure meets the requirements of the urban water supply specification (such as the minimum service pressure ≥0.28 MPa); When more than 5% of the node pressures fail to meet the standard in three consecutive iterations, expand the feasible solution space for pipe diameters (such as adding the DN700 specification option). Finally, output the candidate solutions (the top 3) sorted in ascending order of renovation cost, accompanied by hydraulic stability indicators (such as pressure difference ≤0.15 MPa).

[0027] S5. Conduct a spatial overlay analysis of the optimized pipe diameter combination plan with the road alignment data in the geographic information system, and generate an engineering implementation list containing the pipe section replacement coordinates, new pipeline paths, and valve control instructions in combination with the valve position data, thereby generating a pipe network optimization configuration plan.

[0028] Conduct a spatial overlay analysis of the optimized pipe diameter combination plan (including the start / end coordinates of the pipe section) with the road centerline vector layer in the geographic information system: Establish a buffer zone with a construction safety distance of 1.0 - 1.5 meters outside the road red line, and automatically eliminate the pipeline paths that overlap with existing underground pipe galleries (such as thermal pipelines) by more than 30%. Combine the valve position data (extracted from the pipe network topology data) to establish a node - valve association matrix (identifying the influence range controlled by the valve), and use the breadth - first search algorithm to generate a valve control instruction tree (for example, when a pipe burst occurs, first close the upstream valve and then close the downstream valve). Associate construction feasibility parameters: For pipe sections with a diameter ≥600 mm, verify the path curvature according to the minimum turning radius of the lifting machinery (such as ≥8 meters for an excavator), and automatically insert the coordinates of the transition elbow node when the turning angle > 45°. Finally, generate a structured engineering implementation list: Sort by the unique identification code of the pipe network nodes, and include fields "pipe section number - start coordinate - end coordinate - pipe diameter specification - valve operation sequence - construction machinery model" (for example, a DN400 pipe section uses a 200 - ton crane), output in XML format and ensure data integrity through digital signature.

[0029] This method solves the problem of parameter distortion caused by data islands in traditional modeling through a multi-source spatio-temporal data fusion mechanism; dynamically couples measured data with a hydraulic model, significantly improving the calibration accuracy of key parameters such as friction coefficients; uses a hybrid optimization algorithm to collaboratively process new and old pipe network scenarios, reducing the transformation cost while ensuring water supply reliability; combines spatial analysis with construction constraint verification to greatly reduce the risk of plan changes during the project implementation phase.

[0030] In another technical solution, the establishment of the comprehensive pipe network database in step S1 specifically includes the following steps: Install Internet of Things pressure sensors at the pipe junctions of the target pipe network, and install ultrasonic flow meters at the user water inlet. Transmit the collected hydraulic parameters to the edge computing gateway in real time through the LoRaWAN communication protocol; Convert the pipe network topology data stored in the geographic information system into a vector layer in the WGS84 coordinate system, perform geocoding on the address information in the historical user water consumption record data, and establish a spatial mapping relationship with the pipe network topology nodes; For the real-time hydraulic parameter data stream collected by Internet of Things devices, use the cubic spline interpolation algorithm to reconstruct the data segments with missing timestamps, and perform spatial overlay analysis with the pipe network topology data in the geographic information system to generate a three-dimensional spatio-temporal data cube containing the spatial coordinates of pipe network nodes, pipe section elevation data, and corresponding time series hydraulic parameters; Establish a theoretical calculation model for pipe pressure. When the deviation between the measured value of the sensor and the theoretical value exceeds the safety threshold corresponding to the pipe pressure-bearing grade, automatically trigger data anomaly marking, and use the quartile method to clean the abnormal data, retaining the valid data within the upper and lower quartile intervals; Store the registered pipe network topology data in the PostGIS spatial database, store the time series hydraulic parameter data in the time series database, and establish an association index for the two types of databases by establishing a unique identification code for pipe network nodes, where the unique identification code for pipe network nodes is generated by combining the pipe network number, administrative division code, and spatial coordinate hash value; Set a weekly incremental update period for the geographic information system data, set a 5-minute time window sliding average filtering process for the Internet of Things real-time data stream, and automatically trigger the re-registration calculation of the pipe network topology data and hydraulic parameters when the user water consumption record data is updated.

[0031] Deploy Internet of Things pressure sensors (such as piezoresistive sensors with an optional accuracy of 0.5% FS or 1.0% FS) at the pipe junctions (such as hydraulic sensitive points like tees and elbows) of the target pipe network, and install ultrasonic flow meters (with an accuracy of 1.0% R) at the user water inlet (such as a water meter well). The sensor transmits the pressure value (unit: MPa), flow value (unit: m3 / h) It is transmitted in real time to the edge computing gateway (such as Huawei AR502H). The gateway adds a timestamp (accurate to milliseconds) and a device ID to the original data, and encapsulates it in JSON format and uploads it to the cloud platform through the 4G network. This process solves the problem of insufficient timeliness in traditional manual collection and provides a high-frequency data basis for dynamic modeling (the typical sampling interval is 5 minutes, and the adjustable range is 1 - 15 minutes).

[0032] The pipe network topology data (Shapefile format) in the Geographic Information System (GIS) is converted into point and line layers in the WGS84 coordinate system, and the node elevation is complemented by interpolation of the Digital Elevation Model (DEM) (accuracy ±0.1m). The text addresses (such as "XX Road, XX Number") in the user water consumption records are geocoded to generate longitude and latitude coordinates and establish a mapping relationship with the nearest pipe network node. For the missing values in the Internet of Things data stream (such as signal interruption), the cubic spline interpolation algorithm is used to reconstruct the complete time series (for example, to complement 3 consecutive missing data points). Data quality verification stage: Calculate the theoretical pressure value based on the pipe hydraulic equation (such as Bernoulli equation). When the deviation of the measured value exceeds the pipe pressure-bearing safety threshold (such as PVC pipe ≥0.8MPa, cast iron pipe ≥1.2MPa), it is marked as abnormal, and the quartile method is used for cleaning (removing the data outside the range from Q1 - 1.5IQR to Q3 + 1.5IQR).

[0033] The registered spatial data is stored in the PostGIS database (fields include node ID, coordinates, elevation), and the time series data is stored in the time series database (such as InfluxDB, storing pressure / flow values at 5-minute granularity). Cross-database association indexing is achieved through the unique identification code of the pipe network node (consisting of 16 characters: 6-digit administrative division code + 4-digit pipe network number + 6-digit coordinate hash value). The data update mechanism includes: The GIS data is incrementally updated weekly (only synchronizing the changed area), the Internet of Things data stream is denoised through moving average filtering (window width 5 minutes, adjustable to 3 - 10 minutes), and when the user water consumption is updated, it automatically triggers the re-registration of the topology data and the real-time data (for example, the addition of a new user node causes topology changes).

[0034] Achieve high-precision spatio-temporal alignment of multi-source data, eliminate the fusion error of heterogeneous data; The dynamic data quality monitoring mechanism significantly improves the reliability of the modeling data; The multi-modal storage structure supports the rapid retrieval and analysis of large-scale pipe network data.

[0035] In another technical solution, the following method is used to generate the pipe network optimization configuration plan in step S5: Convert the pipe segment replacement coordinates in the optimized pipe diameter combination plan into the Shapefile format of the Geographic Information System (GIS), conduct buffer analysis with the road centerline vector layer, set a construction safety distance of 1.2 meters outside the road red line, and eliminate the pipe line paths that overlap with the existing underground pipe corridors by more than 30%; establish a node-valve association matrix based on the valve position data in the pipe network topology data, use the breadth-first search algorithm to determine the control influence range of each valve, and generate a control instruction tree diagram including the valve opening and closing sequence and timing; Extract the pipe material attribute data from the comprehensive pipe network database, combine the lifting machinery operation radius parameters at the construction site, conduct curvature radius verification on the pipe segment paths with a pipe diameter ≥ 600 mm, and insert transition elbow nodes when the turning angle exceeds the minimum turning radius of the excavator; associate the new pipe line paths after spatial overlay analysis with the construction feasibility verification results, sort them according to the unique identification code of the pipe network nodes, and output a structured table including the pipe segment number, starting coordinates, ending coordinates, pipe diameter specification, valve control timing, and construction machinery model; When the road alignment data changes, trigger the recalculation of the spatial overlay analysis, and automatically update the coordinate data of the affected pipe segments by comparing the version numbers; use the constraint satisfaction algorithm to verify the temporal and spatial conflicts between the pipe segment replacement and the new construction paths in the project implementation list. When two construction tasks are detected at the same coordinate point, automatically delay the execution time of the non-urgent task and generate a conflict warning report; merge and encode the project implementation list with the valve control instruction tree diagram, generate an XML format document, and ensure data integrity through digital signatures.

[0036] Convert the pipe segment coordinates (WGS84 longitude and latitude) in the optimized pipe diameter combination plan into the Shapefile format of GIS, and conduct buffer analysis with the road centerline layer (the buffer radius is set to 1.2 meters, which can be adjusted to 1.0 - 1.5 meters according to the road grade). The system automatically calculates the spatial overlap degree between the new pipe line paths and the existing underground pipe corridors (such as gas pipelines) (based on vector surface intersection analysis). If the overlapping area > 30% (the threshold can be set to 20% - 40%), then eliminate this path. At the same time, construct a node-valve association matrix in combination with the valve position data (the matrix element is 1 indicating that the valve controls this node), and use the breadth-first search algorithm (BFS) to traverse the pipe network topology in reverse from the pipe burst point to generate a valve closing timing tree (for example, close the upstream main pipeline valve first and delay closing the downstream branch pipeline valve).

[0037] Extract pipe material properties from the integrated pipeline network database (such as the weight of DN600 ductile iron pipe is 12.5 tons / m), and combine with construction site parameters (such as the working radius of a 200-ton crane ≥ 8 meters) to perform path curvature verification: calculate the turning angle of the pipe section (calculate the radius of curvature through three-point coordinates), and when the turning radius < the minimum mechanical turning radius (such as the minimum turning radius of an excavator is 6 meters), automatically insert an elbow node (add a new coordinate point to make the turning angle ≤ 30°). Output a structured engineering list: generate a table sorted by the unique identification code of the pipeline network node, and the fields include the starting point coordinates (longitude, latitude), ending point coordinates, pipe diameter specification (such as DN400), valve operation timing (such as "Valve V001 closes at T + 5 min"), construction machinery model (such as "XCMG 200-ton crane").

[0038] When GIS road data changes (such as adding a median strip), the system locates the affected pipe sections through version number comparison (spatial topology analysis), and automatically triggers coordinate update (recalculate the path when the offset > 0.5 meters). Use a constraint satisfaction algorithm (CSP) to detect construction time and space conflicts: define construction tasks as variables (such as "Replacement of pipe section A: coordinates (X1, Y1), time T1 - T2"), and the constraint condition is that only 1 task is allowed at the same coordinate point within 24 hours. When a conflict is detected (such as coordinate point P is assigned both excavation and lifting tasks at time T1), the system automatically delays non-urgent tasks (the delay duration can be set from 12 to 48 hours) and generates a warning report. The final plan code is an XML document that conforms to the ISO 55000 standard (including a digital signature verification code).

[0039] Realize the deep integration of spatial constraints and construction parameters, greatly reducing the engineering implementation risk; the dynamic conflict detection mechanism effectively avoids construction resource scheduling conflicts; the standardized output format significantly improves the cross-departmental collaboration efficiency.

[0040] In another technical solution, the following method is used to construct a dynamic hydraulic simulation model in step S2: Extract the pipe section end point coordinates according to the pipeline network topology data of the geographic information system, and construct a 0-1 incidence matrix with pipeline network nodes as row vectors and adjacent pipe sections as column vectors, where the flow direction of the pipe section is represented by the positive and negative values determined by the flowmeter data; Divide the pressure data real-time monitored by the pressure sensor into data slices at a 15-minute time granularity, and inject them into the hydraulic calculation model through the EPANET's Toolkit API interface to establish a dynamic mapping relationship between the measured pressure value and the calculated value; Aiming at the problem of the initial value setting error of the pipe section friction coefficient, use the Levenberg-Marquardt algorithm as the iterative optimizer of the nonlinear least squares method, take the root mean square error between the measured pressure data slice and the EPANET calculated value as the objective function, and synchronously optimize the friction coefficient combinations of the adjacent three pipe sections to complete the friction coefficient correction; Substitute the corrected friction coefficient into the EPANET model for transient simulation, extract the flow velocity and head loss data of each pipe segment at different valve openings, construct a pipe network hydraulic state tensor with spatial coordinates as the X-Y axis and time series as the Z axis, and generate a three-dimensional hydraulic characteristic field. Close the pressure regulating valve of the target pipe network during the low water consumption period of each day, obtain the actual pressure distribution data through the SCADA system, and conduct a Kolmogorov-Smirnov test on the simulation results of the three-dimensional hydraulic characteristic field. When the P value is less than 0.05, correct the friction coefficient again.

[0041] Extract the coordinates of the pipe segment endpoints (latitude and longitude values) according to the pipe network topology data of the geographic information system, and construct a node-pipe segment topology relationship matrix: using the pipe network nodes as the row index (such as nodes N001-N500) and the adjacent pipe segments as the column index (such as pipe segments P100-P300), and the matrix element value of 1 indicates a forward connection, and -1 indicates a reverse connection (the flow direction is determined by the data symbol of the flowmeter). The original pressure data (possibly with water hammer noise) monitored by the pressure sensor in real time is segmented into data slices at a fixed time granularity (typically 15 minutes, adjustable from 5 to 30 minutes), and the hydraulic model boundary conditions (such as the water demand at the node) are injected through the Toolkit API interface of EPANET. This process establishes a dynamic mapping relationship between the measured pressure and the calculated value: when the measured pressure at a certain node continuously exceeds the simulated value by more than 0.05 MPa, automatically mark the associated pipe segment as the area where the friction coefficient needs to be optimized.

[0042] For the initial value error of the pipe segment friction coefficient (such as the preset value of 0.018±0.005 for cast iron pipes), use the L-M algorithm (Levenberg-Marquardt algorithm) for nonlinear optimization: take three adjacent pipe segments as the optimization unit (to reduce the risk of local optimum), the objective function is the root mean square error (RMSE) between the measured pressure slice and the EPANET calculated value, and the iteration step size is set to 0.001-0.005. Substitute the corrected friction coefficient into the EPANET transient module, simulate the water flow state under the step change of the valve opening (10% / 30% / 50% / 70% / 90%), extract the flow velocity and head loss data of each pipe segment, and construct a three-dimensional hydraulic characteristic field: the X-Y axis is the spatial coordinate (unit: meter), the Z axis is the time series (unit: hour), and the attribute value includes the pressure / flow velocity matrix (such as 100 nodes × 24 hours × 5 valve states).

[0043] During the low water consumption period of each day (such as 02:00 - 04:00), close 50% of the pressure regulating valves in the target pipe network (select valves on non-main pipes), and collect the actual pressure distribution data through the SCADA system. Conduct the Kolmogorov-Smirnov test (KS test) on the measured data and the simulation values of the three-dimensional hydraulic characteristic field: calculate the cumulative probability difference of the pressure distribution function. When the P value < 0.05 (the significance threshold can be set to 0.01 - 0.1), it is determined that the model is inaccurate. Trigger the iterative correction process: return to the friction coefficient optimization step, and preferentially adjust the parameters of the top 10% of the pipe segments with the largest differences in the KS test until the P value meets the standard or reaches the maximum number of iterations (such as 20 times).

[0044] The dynamic data coupling mechanism significantly improves the real-time adaptability of the hydraulic model. The multi-parameter collaborative correction effectively reduces the calibration error of the friction coefficient, and the iterative mechanism driven by statistical tests ensures the long-term reliability of the model.

[0045] The process of collaborative correction of the friction coefficient includes: Construct a dynamic water consumption pattern library, use a long short-term memory neural network to train the historical record data of user water consumption, generate a water consumption prediction model including weekdays, holidays, and seasonal changes, and output the water consumption change curves of each node in the next 24 hours; discretize the water consumption prediction curve into node water demand boundary conditions at 15-minute intervals, and introduce a dynamic weight factor in the iterative process of the Levenberg-Marquardt algorithm to make the optimization process preferentially match the measured pressure data slices during the peak water consumption period; Establish spatial constraints for the friction coefficient. Determine the change range of the friction coefficient for pipe segments of different materials according to the pipe material attribute data. When the optimized friction coefficient exceeds the expected friction coefficient range of the pipe material, trigger an early warning of the abnormal state of the pipe segment; set a double-threshold convergence criterion. When the change amount of the root mean square error for three consecutive iterations is less than 0.01 m and the change amount of the friction coefficient is less than 0.0001, terminate the optimization and retain the optimal friction coefficient combination; write the finally corrected friction coefficient into the basic input file of the EPANET model, update the pipe material attribute data in the comprehensive pipe network database at the same time, and record the version number and check code.

[0046] Construct a dynamic water usage pattern library: Use a long short-term memory neural network (LSTM) to train the historical data of user water consumption (time span ≥ 1 year). The input features include date type (weekday / holiday), temperature (obtained from the meteorological bureau API), and season identifier (spring / summer / autumn / winter). The output is the change curve of water consumption at each node in the next 24 hours (time resolution 15 minutes), which is discretized into the boundary conditions of water demand at EPANET nodes (unit: L / s). Introduce a dynamic weight factor in the L-M algorithm iteration: Assign a weight of 2.0 - 3.0 (adjustable) to the measured pressure data slices during the peak water consumption period (such as 07:00 - 09:00), and set the weight to 1.0 during off-peak periods, so that the optimization process better fits the actual load changes.

[0047] Set the physical constraint interval of the friction coefficient according to the pipe material property data: Polyvinyl chloride (PVC) pipe: 0.008 - 0.012 (take the lower value for new pipes); Cast iron pipe: 0.015 - 0.020 (service life < 15 years), and relax the upper limit to 0.022 for old cast iron pipes; When the optimized value exceeds the limit, automatically trigger a pipe section anomaly warning (such as internal wall scaling). Use the MPI parallel framework to divide the pipe network into 4 hydraulic zones (grouped according to topological connectivity), and each zone independently runs the L-M optimizer (the number of computing nodes is equal to the number of zones). Boundary data exchange mechanism: Synchronize the pressure data of the boundary junction nodes every 5 iterations (force synchronization when the node pressure difference > 0.03 MPa), to avoid local optimization conflicts.

[0048] Set a double-threshold convergence criterion: Main threshold: The change in RMSE for three consecutive iterations < 0.01 m; Secondary threshold: The change in friction coefficient < 0.0001; Terminate the optimization when either threshold is met. Output a parameter credibility report: Record the correlation coefficient (Pearson coefficient) between the change in friction coefficient of each pipe section and the pressure error reduction rate. When |r| < 0.7, mark it as a "potentially abnormal pipe section" (there may be undetected leakage points). Finally, write the corrected parameters into the EPANET INP base file (ASCII format), synchronously update the pipe material property table in the comprehensive pipe network database, and add new fields "Optimization version number Vxx" and "MD5 checksum" (such as e99a18c4).

[0049] In another technical solution, the multi-stage optimization calculation performed in step S4 includes the following steps: Extract the road width data, underground utility tunnel distribution data, and coordinates of existing buildings in the old pipe network renovation area, establish a two-dimensional rasterized constraint matrix including pipe segment burial depth limits, horizontal offset tolerances, and construction restricted area ranges, and obtain the geographical constraint condition matrix; according to the pipe material attribute data in the comprehensive pipe network database, screen out the set of pipe diameter specifications that meet the construction standards, and perform a Cartesian product operation in combination with the geographical constraint condition matrix to exclude the pipe diameter-path combinations with a distance less than 1.5 meters from underground power pipelines, and construct a feasible solution space for pipe diameters. Implement parallel computing of the hybrid algorithm, embed the velocity update formula of the particle swarm optimization algorithm in the genetic algorithm with an elite retention strategy, divide the feasible solution space of pipe diameters into 4 subspaces, calculate the initial fitness values of each subspace respectively, and retain the solution set with the top 10% of fitness values. Execute the pressure threshold verification, compare the node pressure simulation values output by the dynamic hydraulic simulation model with the minimum service pressure values in the urban water supply specification node by node. When more than 5% of the nodes do not meet the standards within three consecutive iteration cycles, expand the feasible solution space of pipe diameters; arrange the pipe diameter combination schemes that pass the pressure threshold verification in ascending order of renovation cost, and output the top three candidate schemes and their corresponding hydraulic stability indicators to obtain the optimized scheme set.

[0050] Extract the road width data (the default requirement is ≥4 meters to meet the passage of construction machinery, which can be relaxed to 3.5 - 5.0 meters), underground utility tunnel distribution data (such as the coordinates of thermal / electricity pipelines), and coordinates of existing buildings (such as the radius of the historical protection building restricted area ≥5 meters) in the old pipe network renovation area, and construct a two-dimensional rasterized constraint matrix: divide the area into 1m×1m grids (the resolution can be adjusted from 0.5 - 2m), and the grid values are marked as 0 (constructible area) or 1 (restricted area). According to the pipe material attributes of the comprehensive pipe network database, screen out the set of pipe diameter specifications that meet the GB / T 50268 standard (such as ductile iron pipes DN100 - DN600, a total of 12 specifications). Generate the original solution space (pipe diameter specification × path coordinate combination) through the Cartesian product, and automatically exclude the high-risk combinations with a horizontal distance of <1.5 meters from underground power pipelines (the safety threshold can be set from 1.0 - 2.0 meters), and finally form a feasible solution space for pipe diameters (about 10 4 magnitude solutions).

[0051] Embed the velocity update formula of particle swarm optimization (PSO) in the genetic algorithm with an elite retention strategy, including: Genetic algorithm operations: The initial population size is 500, the two-point crossover probability is 0.7 - 0.9, and the mutation probability is 0.1 - 0.3; PSO fusion mechanism: 10% of the optimal solutions of each generation are injected into the particle swarm, and the inertia weight of 0.6 - 0.9 is introduced into the particle velocity update formula; The feasible solution space is divided into 4 sub-spaces (partitioned by the topological connectivity of the pipe network), and the fitness values (including the transformation cost and the reduction in leakage) are calculated independently for each sub-space. In each iteration, the top 10% of the solution sets in terms of fitness (about 50 solutions) are retained, and the remaining solutions are eliminated. The dynamic hydraulic model verifies the node pressures: EPANET is called to simulate the water flow states corresponding to each solution, and the node pressures are compared with the minimum value specified in CJJ 207 (such as 0.28 MPa). When more than 5% of the nodes do not meet the standard, the solution space is expanded (such as adding a DN700 specification).

[0052] The solution sets that pass the pressure verification are sorted in ascending order of the transformation cost, and the top three candidate solutions are output. The calculation of the hydraulic stability index includes: the pressure range (the difference between the highest and lowest node pressures ≤ 0.15 MPa), the variance of pressure fluctuations (the variance of the 24-hour simulation value ≤ 0.01), and the identification of bottleneck pipe sections (the proportion of pipe sections with a flow velocity > 2.5 m / s < 5%); the final solution is accompanied by a construction risk assessment report (such as the path curvature analysis for a DN600 pipe section that requires a 200-ton crane operation).

[0053] Through the deep integration of geographical constraints and specification standards, illegal solutions are avoided at the source; a hybrid algorithm is used to synergistically improve the optimization efficiency of complex pipe networks, and multi-dimensional stability indicators ensure the reliability of the water supply system.

[0054] The construction of the feasible solution space for pipe diameters includes the following steps: A multi-objective conflict resolution model is established, and the Pareto front analysis method is used to handle the trade-off relationship between the transformation cost and hydraulic stability in pipe diameter selection. The condition for screening valid solutions is set that the increase rate of the transformation cost does not exceed 1.5 times the increase rate of hydraulic stability; a random forest classifier is trained using historical project data to predict the construction feasibility probability of each pipe diameter-path combination, and when the predicted value is lower than 0.7, this combination is automatically excluded; A dynamic solution space index is generated. The feasible solution space for pipe diameters is established with a B+ tree index structure encoded by administrative divisions, and the solution space data is retrieved and updated in combination with the unique identification code of the pipe network nodes; the principal component analysis method is used to extract the key feature vectors of the pipe diameter combination scheme, reducing the solution space dimension from the original number of pipe sections n to the log2(n) level; 5% of the feasible solutions are randomly selected for on-site inspection and verification. When the non-pass rate of the verification exceeds 20%, the regeneration of the geographical constraint condition matrix is triggered.

[0055] Build a Pareto front analysis model: The horizontal axis represents the pipe network transformation cost (in ten thousand yuan), and the vertical axis represents the hydraulic stability index (dimensionless score 0 - 1). The frontier solution set is screened through non-dominated sorting (about 20% of the total number of solutions). Set the effective solution screening condition: If a certain solution improves the stability by ΔS% and the transformation cost increases by ΔC%, then it is necessary to satisfy ΔC% / ΔS% ≤ 1.5 (the threshold can be adjusted from 1.2 to 2.0). Use the historical engineering database (including more than 500 cases) to train a random forest classifier: The input features include pipe diameter specifications, soil types, road grades, etc.; the output is the construction feasibility probability (0 - 1). When the predicted probability < 0.7 (the threshold can be set from 0.6 to 0.8), this combination is automatically excluded (such as the laying risk of DN500 pipes in soft soil subgrades).

[0056] Use principal component analysis (PCA) for dimensionality reduction: Extract 5 principal components of the pipe diameter combination scheme (cumulative contribution rate > 85%), and reduce the dimension of the solution space from the original number of pipe segments n (such as 200 dimensions) to the log2(n) level (about 8 dimensions). Establish a B+ tree index structure: Use the administrative division code (such as 110101) as the root node key value, and the leaf node stores the unique identification code of the pipe network node (16-bit code) and the corresponding solution space pointer. Index update mechanism: When a new user node is added, the solution space branch under this administrative division is automatically expanded (such as adding a DN400 pipe segment combination).

[0057] Randomly select 5% of the feasible solutions (about 50 solutions) for on-site investigation: The verification content includes the actual location of underground obstacles (such as unmarked optical cables) and temporary construction conditions (such as soil bearing capacity in the rainy season). When the non-passing rate of verification > 20% (the threshold can be set from 15% to 25%), trigger the regeneration of the geographical constraint condition matrix: Adjust the no-go zone radius (such as expanding the building no-go zone from 3 meters to 4 meters) and supplement newly discovered pipe gallery data. The verification results feed back to the machine learning model to incrementally update the random forest classifier.

[0058] Significantly reduce the risk of scheme implementation through an intelligent screening mechanism. The dimensionality reduction index can improve the retrieval efficiency of a large-scale solution space, and the on-site feedback closed-loop optimizes the geographical constraint accuracy.

[0059] The implementation of the hybrid algorithm for parallel computing includes the following specific steps: Construct a distributed computing architecture, use the MPI parallel computing framework to allocate 4 sub-spaces to different computing nodes, and each node independently runs the hybrid optimization algorithm and synchronizes the optimal solution data every 20 minutes; Dynamically adjust the crossover probability of the genetic algorithm according to the dispersion degree of the sub-space solution set. When the standard deviation of the solution set exceeds the set threshold, reduce the crossover probability from 0.8 to 0.6; Monitor the change rate of the fitness value of each subspace. When the change rate is less than 0.1% within three consecutive calculation cycles, terminate the iterative calculation of this subspace in advance; perform non-dominated sorting on the optimal solutions of the four subspaces, and use the crowding comparison operator to select evenly distributed Pareto optimal solutions to form the final optimization solution set; when the solution set of a certain subspace is empty due to geographical constraint conflicts, relax the buried depth limit of the pipeline section in this area by 10% and restart the calculation process.

[0060] Use the MPI parallel framework (such as OpenMPI) to allocate 4 subspaces to independent computing nodes (each node is configured with an 8-core CPU + 32GB of memory). Data synchronization mechanism between nodes: Exchange the optimal solutions of each subspace every 20 minutes (adjustable from 10 to 30 minutes), compare the differences of the solutions through hash values (such as SHA-256 checksum), and the different solution sets trigger global non-dominated sorting and recombination; Dynamic adjustment of genetic algorithm parameters: Monitor the standard deviation σ of the subspace solution set. When σ > 0.15 (the threshold can be set from 0.1 to 0.2), reduce the crossover probability from 0.8 to 0.6 and increase the mutation probability from 0.2 to 0.3 to enhance the population diversity.

[0061] Monitor the change rate δ of the fitness value of the subspace (δ = |f t -f t-1 | / f t-1 ): When δ < 0.1% (the threshold can be set from 0.05% to 0.2%) within three consecutive calculation cycles (each cycle is 20 minutes), terminate the calculation of this subspace in advance. Fuse the Pareto optimal solutions (about 200) output by the four subspaces: Perform non-dominated sorting and grading (Rank1 is the optimal solution set), use the crowding comparison operator (CrowdingDistance), and select 50 evenly distributed solutions to form the final solution set; when the solution set of a certain subspace is empty due to geographical constraint conflicts (such as the entire area being a construction restricted area), automatically relax the buried depth limit of the pipeline section by 10% (such as from the original limit of 2.0 meters to 2.2 meters) and restart the calculation.

[0062] Establish an iterative repair mechanism: When an empty solution subspace is detected, the system automatically executes: Locate the conflict area (such as the coordinate range X1 - Y1), relax the constraint parameters (buried depth / horizontal offset + 10%), and record the number of relaxations (if it exceeds 3 times, mark it as a "high conflict area"); The final output optimization solution set is accompanied by a conflict handling log (such as "The buried depth in area A is relaxed to 2.2 meters"), guiding the priority order of project implementation.

[0063] Through dynamic parameter adjustment, it can effectively avoid the local optimal trap. The early stopping mechanism significantly saves computing resources, and the conflict adaptive repair improves the feasibility of the solution.

[0064] In another technical solution, the specific steps of constructing the dynamic hydraulic simulation model in step S2 are as follows: Establish a dynamic data fusion interface, convert the pipe network topology data of the geographic information system into the GraphML format with topology verification, where the pipe segment connection relationship is stored through an adjacency list, and the node elevation data is interpolated and complemented through a digital elevation model; implement real-time data coupling, perform Kalman filtering on the original pressure data stream collected by the pressure sensor to eliminate the transient fluctuations caused by the water hammer effect, generate the smoothed pressure time series data, and inject the hydraulic model boundary conditions through the INP file interface of EPANET; Execute multi-parameter joint calibration, adopt the nonlinear least squares method with constraint conditions, use the pipe segment friction coefficient and the basic water demand of the node as the synchronous optimization variables, and set the physical constraint interval of the initial value of the pipe friction coefficient; Construct a three-dimensional hydraulic response surface, simulate the step change process of the valve opening from 10% to 90% through the transient analysis module of EPANET, record the gradient matrix of the flow velocity of each pipe segment with the change of the valve state, and generate a three-dimensional hydraulic characteristic space composed of the pipe diameter, elevation, and valve opening; During the stable water use period in the middle of the night every day, close 50% of the regulating valves in the target area, collect the actual pressure distribution data through the SCADA system, and perform dynamic time warping matching with the simulation values in the three-dimensional hydraulic characteristic space. When the mean absolute error exceeds 0.05 MPa, trigger the iterative correction of the friction coefficient.

[0065] Convert the pipe network topology data of the geographic information system into the GraphML format (an XML-based graph structure data format), ensure the integrity of the pipe segment connection relationship through topology verification: use an adjacency list to store the connection relationship between nodes (such as node A connecting pipe segments P1 and P2), and the node elevation data is interpolated and complemented through a digital elevation model (DEM) (accuracy ±0.5 meters). Perform Kalman filtering and noise reduction on the original pressure data stream collected by the pressure sensor: establish a state equation (including the pressure value and its first derivative), the observation equation integrates the pressure readings of the SCADA system, and the filter gain coefficient is set to 0.6 - 0.9 (adjustable) to effectively eliminate the transient fluctuations caused by the water hammer effect (such as pressure mutation > 0.15 MPa / second). The smoothed pressure time series data (sampling interval 15 minutes) is injected into the hydraulic model boundary conditions through the INP file interface of EPANET to realize the dynamic coupling of real-time data and the static model.

[0066] Execute multi-parameter joint calibration: use the pipe segment friction coefficient and the basic water demand of the node as the synchronous optimization variables, and adopt the nonlinear least squares method with constraints. Set the physical constraint interval: Friction coefficient of cast iron pipe: 0.016 - 0.020 (initial value 0.018 ± 0.002); Friction coefficient of polyethylene pipe: 0.010 - 0.012 (initial value 0.011 ± 0.001); Simulate the step change of valve opening (10%, 30%, 50%, 70%, 90% opening) through the EPANET transient analysis module, and record the gradient matrix of the flow velocity of each pipe section changing with the valve state (for example, the flow velocity of the DN400 pipe is 1.2 m / s at 50% opening and 1.8 m / s at 70% opening). Construct a three-dimensional hydraulic response surface: the X-axis is the pipe diameter (DN100 - DN600), the Y-axis is the elevation (0 - 50 meters), the Z-axis is the valve opening (10% - 90%), and the surface value represents the pressure distribution characteristics (unit: MPa).

[0067] Close 50% of the regulating valves in the target area (preferably branch valves) during the stable daily water use period (02:00 - 04:00), and collect the actual pressure distribution data through the SCADA system (sampling density ≥ 3 points / km²). Use the dynamic time warping (DTW) algorithm to match the measured data with the simulated values in the three-dimensional hydraulic characteristic space: calculate the similarity of the pressure curve shape. When the mean absolute error (MAE) exceeds 0.05 MPa (the threshold can be set at 0.03 - 0.08 MPa), automatically trigger the iterative correction process of the friction coefficient. The correction priority is sorted according to the error size: optimize the pipe sections with MAE > 0.1 MPa first, and regenerate the response surface after each correction.

[0068] The availability of pressure data can be significantly improved through dynamic filtering, the physical rationality of the model is enhanced by multi-parameter joint calibration, and the sensitive area of valve regulation is intuitively characterized by the response surface.

[0069] Performing multi-parameter joint calibration includes the following steps: Construct a dynamic decomposition model of water use patterns, and use the variational mode decomposition algorithm to decompose the historical record data of user water consumption into 6 intrinsic mode function components, and extract the time-frequency characteristics of the daily cycle component and the random fluctuation component; Introduce a time-varying weight factor into the objective function of the nonlinear least squares method to increase the weight of the measured data in the corresponding period of the daily cycle component to 2.5 times that of the random fluctuation component; according to the service life data in the pipe material attribute database, set the upper limit of the friction coefficient optimization for cast iron pipe segments over 20 years to 0.022, and set the lower limit of the optimization for polyethylene pipe segments replaced within 5 years to 0.009; use the MPI parallel computing framework to divide the pipe network into 4 hydraulic zones, and each zone independently runs the Levenberg-Marquardt optimizer to achieve global convergence through the exchange of boundary node pressure data; generate a parameter credibility report, record the correlation coefficient between the change in the friction coefficient and the pressure error decline rate in each iteration, and mark the pipe segment as a potential abnormal pipe segment and output the spatial coordinates when the absolute value of the correlation coefficient is less than 0.7.

[0070] Construct a dynamic decomposition model of water use patterns: use the variational mode decomposition (VMD) algorithm to decompose the historical user water consumption data (time series length ≥ 365 days) into 6 intrinsic mode function (IMF) components (the number of components can be adjusted from 5 to 8). Extract key feature components: IMF1: daily cycle component (frequency 0.04 - 0.06Hz); IMF2: random fluctuation component (frequency > 0.1Hz); IMF3 - 6: seasonal / holiday components; calculate the instantaneous frequency of each component through Hilbert transform and construct a time-frequency feature matrix (24 hours × frequency intensity). Introduce a time-varying weight factor into the objective function of the nonlinear least squares method: assign a weight of 2.5 (adjustable from 2.0 to 3.0 times) to the measured data in the period dominated by the daily cycle component (such as 07:00 - 09:00), and set the weight of the random fluctuation period to 1.0.

[0071] Strengthen the constraints based on the service life data in the pipe material attribute database: For cast iron pipes with service life > 20 years: the upper limit of friction coefficient optimization is 0.022 (standard upper limit 0.020); for polyethylene pipes replaced within < 5 years: the lower limit of optimization is 0.009 (standard lower limit 0.010).

[0072] Use the MPI parallel framework (such as MPICH2) to divide the pipe network into 4 hydraulic zones (grouped according to the principle of hydraulic similarity), and each zone is assigned an independent computing node to run the Levenberg-Marquardt optimizer. Boundary data synchronization mechanism: Exchange the pressure data of the boundary nodes between zones every 10 iterations (force synchronization when the node pressure difference > 0.02MPa), and the global convergence condition is that the boundary pressure difference of all zones < 0.01MPa.

[0073] Generate a credibility report for parameters: Record the correlation coefficient (Pearson coefficient) between the change in friction coefficient ΔC and the pressure error decline rate ΔE in each iteration. Set the credibility threshold: |r|≥0.7: The optimization is effective (green mark); 0.5≤|r|<0.7: The optimization is questionable (yellow mark); |r|<0.5: The optimization is ineffective (red mark). When |r|<0.7, mark it as a potentially abnormal pipe section and output the spatial coordinates and the type of abnormality (such as "Coordinates (116.3E, 39.9N): |r| = 0.62, suspected inner wall corrosion"). Finally, generate an optimized parameter version file (including timestamp and digital signature) and automatically update it to the integrated pipe network database.

[0074] Precisely capture the characteristics of water usage patterns through a time-varying weight mechanism, enhance the adaptability of old pipe networks through age-differentiated constraints, and achieve early warning of hidden defects through credibility assessment.

[0075] It should be noted that although the above steps are described in a specific order, it does not mean that the steps must be executed in the above specific order. In fact, some of these steps can be executed concurrently or even in a different order, as long as the required functions can be achieved. The number of devices and the processing scale described here are used to simplify the description of the present invention, and it is obvious to those skilled in the art for the application, modification, and variation of the present invention.

[0076] Although the embodiments of the present invention have been disclosed as above, it is not limited to only the applications listed in the specification and the embodiments. It can be fully applied to various fields suitable for the present invention. For those familiar with the field, additional modifications can be easily made. Therefore, without departing from the general concept defined by the claims and the equivalent scope, the present invention is not limited to specific details and the illustrated and described examples here.

Claims

1. A modeling and analysis method for optimizing urban water supply networks, characterized in that, It includes the following steps: S1. Obtain multi-source heterogeneous data, including: collecting hydraulic parameters of pressure sensors and flow meters in the target pipe network through Internet of Things devices, and synchronously obtaining pipe network topology data, pipe material attribute data, and historical user water consumption record data in the geographic information system; registering the obtained multi-source heterogeneous data according to a unified spatio-temporal coordinate system and storing it in a relational database to establish a comprehensive pipe network database; S2. Based on the pipe network topology data, establish a node-segment topology relationship matrix, couple the real-time monitoring data of pressure sensors with the EPANET hydraulic calculation model, correct the pipe segment friction coefficient through the nonlinear least squares method, generate a three-dimensional pipe network hydraulic model including pipe diameter, elevation, and valve status, and construct a dynamic hydraulic simulation model; S3. For the renovation requirements of the old pipe network, establish a fitness function with the pipe network renovation cost and the reduction of leakage water volume as dual objectives, and use a genetic algorithm with an elite retention strategy to sort the priority of pipe segment replacement; for the new pipe network planning requirements, establish a particle swarm optimization algorithm with construction cost and hydraulic stability as constraint conditions to obtain a hybrid optimization algorithm framework; S4. Perform multi-stage optimization calculations, input the pipe material attribute data in the comprehensive pipe network database into the dynamic hydraulic simulation model for initial hydraulic balance calculations, generate a feasible solution space for pipe diameters according to the geographical constraint conditions in the renovation area of the old pipe network, and iteratively calculate through the hybrid optimization algorithm framework to obtain a pipe diameter combination plan that meets the node pressure threshold; S5. Perform a spatial overlay analysis on the optimized pipe diameter combination plan and the road alignment data in the geographic information system, and generate an engineering implementation list including pipe segment replacement coordinates, new pipeline paths, and valve control instructions in combination with the valve position data to generate a pipe network optimization configuration plan.

2. The modeling and analysis method for optimizing urban water supply networks according to claim 1, characterized in that The specific steps for establishing the comprehensive pipe network database in step S1 include the following steps: Install Internet of Things pressure sensors at the intersections of the pipes in the target pipe network, install ultrasonic flow meters at the user water inlet ends, and transmit the collected hydraulic parameters to the edge computing gateway in real time through the LoRaWAN communication protocol; Convert the pipe network topology data stored in the geographic information system into a vector layer in the WGS84 coordinate system, perform geocoding on the address information in the historical user water consumption record data, and establish a spatial mapping relationship with the pipe network topology nodes; For the real-time hydraulic parameter data stream collected by the Internet of Things devices, use the cubic spline interpolation algorithm to reconstruct the data segment with missing timestamps, and perform a spatial overlay analysis with the pipe network topology data in the geographic information system to generate a three-dimensional spatio-temporal data cube including pipe network node spatial coordinates, pipe segment elevation data, and corresponding time series hydraulic parameters; Establish a calculation model for the theoretical value of pipeline pressure. When the deviation between the measured value of the sensor and the theoretical value exceeds the safety threshold corresponding to the pipeline pressure-bearing grade, automatically trigger data anomaly marking, and use the quartile method to clean the abnormal data, and retain the valid data within the upper and lower quartile intervals; Store the registered pipe network topology data in the PostGIS spatial database, and store the time series hydraulic parameter data in the time series database. Establish an association index for the two databases by creating a unique identification code for pipe network nodes, where the unique identification code for pipe network nodes is generated by combining the pipe network number, administrative division code, and spatial coordinate hash value; Set the weekly incremental update period for geographic information system data, set a sliding average filtering process with a 5-minute time window for the real-time Internet of Things data stream, and automatically trigger the re-registration calculation of pipe network topology data and hydraulic parameters when the user water consumption record data is updated.

3. The modeling and analysis method for optimizing urban water supply networks according to claim 1, wherein In step S5, the following method is used to generate the pipe network optimization configuration plan: Convert the pipe segment replacement coordinates in the optimized pipe diameter combination plan into the Shapefile format of the geographic information system, perform buffer analysis with the road centerline vector layer, set 1.2 meters outside the road red line as the construction safety distance, and eliminate the pipe line paths that overlap with the existing underground pipe gallery by more than 30%; establish a node-valve association matrix according to the valve position data in the pipe network topology data, and use the breadth-first search algorithm to determine the control influence range of each valve, and generate a control instruction tree diagram including the valve opening and closing sequence and time series; Extract the pipe material attribute data from the comprehensive pipe network database, combine the lifting machinery operation radius parameters at the construction site, check the curvature radius of the pipe segment paths with a pipe diameter ≥ 600 mm, and insert transition elbow nodes when the turning angle exceeds the minimum turning radius of the excavator; associate the new pipe line paths after spatial overlay analysis with the construction feasibility check results, sort them according to the unique identification code of the pipe network nodes, and output a structured table including the pipe segment number, start point coordinates, end point coordinates, pipe diameter specifications, valve control time series, and construction machinery models; When the road alignment data changes, trigger the recalculation of the spatial overlay analysis, and automatically update the coordinate data of the affected pipe segments by comparing the version numbers; use the constraint satisfaction algorithm to check the spatio-temporal conflicts of the pipe segment replacement and new construction paths in the project implementation list. When two construction tasks are detected at the same coordinate point, automatically postpone the execution time of the non-urgent task and generate a conflict warning report; merge and encode the project implementation list and the valve control instruction tree diagram, generate an XML format document, and ensure data integrity through digital signatures.

4. The modeling and analysis method for optimizing urban water supply networks according to claim 1, characterized in that In step S2, the following method is used to construct a dynamic hydraulic simulation model: Extract the pipe segment end point coordinates according to the pipe network topology data of the geographic information system, and construct a 0-1 association matrix with the pipe network nodes as row vectors and adjacent pipe segments as column vectors, where the pipe segment flow direction is represented by the positive and negative values determined by the flow meter data; Divide the pressure data real-time monitored by the pressure sensor into data slices at a 15-minute time granularity, inject them into the hydraulic calculation model through the EPANET Toolkit API interface, and establish a dynamic mapping relationship between the measured pressure value and the calculated value; Aiming at the problem of the initial value setting error of the pipe section friction coefficient, the Levenberg-Marquardt algorithm is used as the iterative optimizer of the nonlinear least squares method. Taking the root mean square error between the measured pressure data slices and the EPANET calculated values as the objective function, the friction coefficient combinations of three adjacent pipe sections are synchronously optimized to complete the friction coefficient correction. Substitute the corrected friction coefficient into the EPANET model for transient simulation, extract the flow velocity and head loss data of each pipe section at different valve openings, construct a pipe network hydraulic state tensor with spatial coordinates as the X-Y axis and time series as the Z axis, and generate a three-dimensional hydraulic characteristic field. During the low water consumption period of each day, close the pressure regulating valve of the target pipe network, obtain the actual pressure distribution data through the SCADA system, and conduct a Kolmogorov-Smirnov test on the simulation results of the three-dimensional hydraulic characteristic field. When the P value is less than 0.05, re-correct the friction coefficient.

5. The modeling and analysis method for optimizing urban water supply networks according to claim 4, characterized in that The process of collaborative correction of the friction coefficient includes: Construct a dynamic water use pattern library, use a long short-term memory neural network to train the historical record data of user water consumption, generate a water consumption prediction model including weekdays, holidays, and seasonal changes, and output the water consumption change curves of each node in the next 24 hours; discretize the water consumption prediction curve into node water demand boundary conditions at 15-minute intervals, and introduce a dynamic weight factor in the iterative process of the Levenberg-Marquardt algorithm to make the optimization process preferentially match the measured pressure data slices during the peak water use period. Establish spatial constraints for the friction coefficient. Determine the friction coefficient change range of pipe sections with different materials according to the pipe material attribute data. When the optimized friction coefficient exceeds the expected friction coefficient range of the pipe material, trigger an abnormal state warning for the pipe section; set a double-threshold convergence criterion, and terminate the optimization when the change amount of the root mean square error in three consecutive iterations is less than 0.01 m and the change amount of the friction coefficient is less than 0.0001, and retain the optimal friction coefficient combination; write the finally corrected friction coefficient into the basic input file of the EPANET model, update the pipe material attribute data in the comprehensive pipe network database at the same time, and record the version number and check code.

6. The modeling and analysis method for optimizing urban water supply networks according to claim 1, characterized in that The multi-stage optimization calculation performed in step S4 includes the following steps: Extract the road width data, underground utility tunnel distribution data, and existing building coordinate data in the old pipe network renovation area, establish a two-dimensional rasterized constraint matrix including pipe section burial depth limits, horizontal offset tolerances, and construction restricted area ranges, and obtain a geographical constraint condition matrix; according to the pipe material attribute data in the comprehensive pipe network database, screen out the set of pipe diameter specifications that meet the construction standards, perform a Cartesian product operation in combination with the geographical constraint condition matrix, and exclude the pipe diameter-path combinations with a distance less than 1.5 meters from the underground power pipeline to construct a feasible solution space for the pipe diameter. Implement hybrid algorithm parallel computing. Embed the velocity update formula of the particle swarm optimization algorithm into the genetic algorithm with an elite retention strategy, divide the feasible solution space of the pipe diameter into 4 sub-spaces, calculate the initial fitness values of each sub-space respectively, and retain the solution set with the top 10% fitness values. Execute the pressure threshold verification, compare the node pressure simulation values output by the dynamic hydraulic simulation model with the minimum service pressure values in the urban water supply specification node by node. When more than 5% of the nodes do not meet the standards within three consecutive iteration cycles, expand the feasible solution space of the pipe diameter; arrange the pipe diameter combination schemes that pass the pressure threshold verification in ascending order of the transformation cost, output the top three candidate schemes and their corresponding hydraulic stability indicators, and obtain the optimized scheme set.

7. The modeling and analysis method for optimizing urban water supply networks according to claim 6, characterized in that, Constructing the feasible solution space of the pipe diameter includes the following steps: Establish a multi-objective conflict resolution model, adopt the Pareto front analysis method to handle the trade-off relationship between the transformation cost and hydraulic stability in pipe diameter selection, and set that the increase rate of the transformation cost does not exceed 1.5 times the increase rate of hydraulic stability as the effective solution screening condition; use historical project data to train a random forest classifier to predict the construction feasibility probability of each pipe diameter-path combination, and automatically exclude the combination when the predicted value is lower than 0.

7. Generate a dynamic solution space index, establish a B+ tree index structure for the feasible solution space of the pipe diameter according to the administrative division code, and combine the unique identification code of the pipe network node to retrieve and update the solution space data; adopt the principal component analysis method to extract the key feature vectors of the pipe diameter combination scheme, and reduce the solution space dimension from the original number of pipe segments n to the log2(n) level; randomly select 5% of the feasible solutions for on-site inspection verification, and trigger the regeneration of the geographic constraint condition matrix when the non-pass rate of the verification exceeds 20%.

8. The modeling and analysis method for optimizing urban water supply pipe networks according to claim 7, characterized in that Implementing the parallel calculation of the hybrid algorithm includes the following specific steps: Construct a distributed computing architecture, adopt the MPI parallel computing framework to allocate 4 sub-spaces to different computing nodes, each node independently runs the hybrid optimization algorithm and synchronizes the optimal solution data every 20 minutes; dynamically adjust the crossover probability of the genetic algorithm according to the dispersion degree of the sub-space solution set, and reduce the crossover probability from 0.8 to 0.6 when the standard deviation of the solution set exceeds the set threshold. Monitor the change rate of the fitness value of each sub-space. When the change rate is less than 0.1% within three consecutive calculation cycles, terminate the iterative calculation of the sub-space in advance; perform non-dominated sorting on the optimal solutions of the four sub-spaces, and adopt the crowding degree comparison operator to select the evenly distributed Pareto optimal solutions to form the final optimized scheme set; when the solution set of a certain sub-space is empty due to geographical constraint conflicts, relax the pipe section burial depth limit in this area by 10% and restart the calculation process.

9. The modeling and analysis method for optimizing urban water supply networks according to claim 1, characterized in that The specific steps for constructing the dynamic hydraulic simulation model in step S2 are as follows: Establish a dynamic data fusion interface, convert the pipe network topology data of the geographic information system into the GraphML format with topology verification, where the pipe section connection relationship is stored through the adjacency list, and the node elevation data is interpolated and complemented through the digital elevation model; implement real-time data coupling, perform Kalman filtering on the original pressure data stream collected by the pressure sensor to eliminate the transient fluctuations caused by the water hammer effect, generate the smoothed pressure time series data, and inject the hydraulic model boundary conditions through the INP file interface of EPANET. Execute multi-parameter joint calibration, adopt the non-linear least squares method with constraint conditions, take the pipe section friction coefficient and the basic water demand of nodes as the synchronous optimization variables, and set the physical constraint interval of the initial value of the pipe friction coefficient; Construct a three-dimensional hydraulic response surface, simulate the step change process of the valve opening from 10% to 90% through the transient analysis module of EPANET, record the gradient matrix of the flow velocity of each pipe section with the change of the valve state, and generate a three-dimensional hydraulic characteristic space composed of pipe diameter, elevation and valve opening; During the stable water use period in the late night every day, close 50% of the regulating valves in the target area, collect the actual pressure distribution data through the SCADA system, and perform dynamic time warping matching with the simulation values in the three-dimensional hydraulic characteristic space. When the mean absolute error exceeds 0.05 MPa, trigger the iterative correction of the friction coefficient.

10. The modeling and analysis method for optimizing urban water supply networks according to claim 9, characterized in that The execution of multi-parameter joint calibration includes the following steps: Construct a dynamic decomposition model of the water use pattern, and use the variational mode decomposition algorithm to decompose the historical record data of the user water consumption into 6 intrinsic mode function components, and extract the time-frequency characteristics of the daily cycle component and the random fluctuation component; Introduce a time-varying weight factor into the objective function of the non-linear least squares method to increase the weight of the measured data in the corresponding period of the daily cycle component to 2.5 times that of the random fluctuation component; according to the service life data in the pipe material attribute database, set the upper limit of the friction coefficient optimization for cast iron pipe sections over 20 years to 0.022, and set the lower limit of the optimization for polyethylene pipe sections replaced within 5 years to 0.009; use the MPI parallel computing framework to divide the pipe network into 4 hydraulic partitions, and each partition runs the Levenberg-Marquardt optimizer independently, and achieve global convergence through the exchange of boundary node pressure data; generate a parameter credibility report, record the correlation coefficient between the change amount of the friction coefficient and the pressure error decline rate in each iteration, and mark the pipe section as a potential abnormal pipe section and output the spatial coordinates when the absolute value of the correlation coefficient is less than 0.7.

Images