A regional water supply network leakage analysis and early warning system based on the Internet of Things

Abstract

This invention relates to the field of analysis and early warning, and discloses an IoT-based regional water supply network leakage analysis and early warning system. The system includes a partitioning module, a fusion analysis module, and an early warning module. The system collects data through sensor arrays deployed in the network partitions. The fusion analysis module executes a negative pressure wave localization algorithm and an acoustic correlation analysis algorithm in parallel, generating first and second localization results respectively. By calculating the Euclidean distance between the two results and comparing it with a preset tolerance, it determines whether to output localization information or trigger model reverse verification. During verification, a dynamic hydraulic model takes real-time data as input, iteratively fits and outputs a virtual leak location, and outputs the final result after a second distance verification. The early warning module generates tiered early warning information based on the analysis results and displays it visually. This invention, by combining multi-algorithm fusion verification with hydraulic model reverse verification, significantly improves the accuracy and reliability of leakage localization and effectively reduces the false alarm rate.

Description

Technical Field

[0001] This invention relates to the field of analysis and early warning, and specifically to an Internet of Things-based regional water supply network leakage analysis and early warning system. Background Technology

[0002] With the rapid development of my country's high-speed railway network, modern high-speed railway stations have become comprehensive transportation hubs integrating transportation, commerce, and services. Their internal water supply network systems are massive and complex, covering not only sanitary ware and fire-fighting facilities in numerous public areas such as waiting halls and commercial areas, but also key equipment areas such as data centers and air conditioning / refrigeration rooms that support the station's core operations. However, this network system is usually concealed in ceilings, utility tunnels, or underground spaces, making its operational status difficult to monitor directly.

[0003] Currently, for leak detection in regional water supply networks, the industry commonly uses methods such as threshold alarms from independent pressure and flow sensors or periodic manual inspections. These traditional methods have significant limitations:

[0004] First, threshold alarms are highly delayed, typically only being triggered after leakage has occurred and continued for a period of time, until the water volume or pressure changes to a significant degree, thus failing to provide early warning.

[0005] Secondly, manual inspections have long cycles and many blind spots, making it difficult to effectively cover pipelines buried inside structures or in critical computer rooms.

[0006] Furthermore, existing positioning technologies, such as the single negative pressure wave method or acoustic correlation analysis method, suffer significant losses in positioning accuracy and reliability under the unique complex pipeline structure of high-speed railway stations and the high-intensity environmental noise interference such as crowd noise and equipment operation.

[0007] The aforementioned problems are amplified dramatically in the specific context of high-speed rail stations. If leaks in the waiting hall pipes are not detected promptly, they can lead to water accumulation on the floor, causing serious safety hazards such as passenger slips and falls, and electrical short circuits, severely impacting station operations and the image of public services. Even minor leaks in the equipment room area are more insidious and destructive; long-term leakage not only wastes precious water resources but also jeopardizes sophisticated and expensive communication equipment and power control systems. Summary of the Invention

[0008] The purpose of this invention is to provide an Internet of Things-based regional water supply network leakage analysis and early warning system to solve the above-mentioned technical problems.

[0009] The objective of this invention can be achieved through the following technical solutions:

[0010] An IoT-based regional water supply network leakage analysis and early warning system includes:

[0011] The partitioning module is used to divide the water supply network into multiple independent partitions and to deploy IoT sensor arrays at the inlet, outlet and key internal nodes of each partition.

[0012] The fusion analysis module runs on a cloud platform and is communicatively connected to the sensor array.

[0013] The fusion analysis module is used to execute the verification process, which outputs the leakage location result through the following steps:

[0014] The negative pressure wave localization algorithm based on pressure transients and the acoustic correlation analysis algorithm based on pipeline noise are invoked in parallel to generate the first localization result and the second localization result, respectively.

[0015] Calculate the first Euclidean distance between the first positioning result and the second positioning result;

[0016] Based on the comparison result between the first Euclidean distance and the first preset distance tolerance, determine whether to output the final location information of the leak point or trigger the subsequent verification process.

[0017] The early warning module receives the output from the cloud platform analysis module, generates tiered early warning information, and displays it on a visualization map based on partitions and nodes.

[0018] In a further embodiment, the sensor includes a pressure sensor, a flow meter, and a noise sensor;

[0019] The comparison process between the first Euclidean distance and the first preset distance tolerance is as follows:

[0020] If the first Euclidean distance is less than the first preset distance tolerance, it is determined that the positioning is consistent, and the positioning information with the geometric center of the two positioning results as the final leak point position is output.

[0021] If the first Euclidean distance is greater than or equal to the first preset distance tolerance, it is determined that the positioning is inconsistent and the subsequent verification process is triggered.

[0022] A further solution is that the negative pressure wave localization algorithm captures the instantaneous negative pressure wave generated when a pipeline leaks, and uses the time difference between the time the pressure sensors deployed at both ends of the pipeline receive the negative pressure wave to perform preliminary localization of the leak point and obtain a first localization result.

[0023] The acoustic correlation analysis algorithm analyzes the leakage noise signals received by two noise sensors deployed at two different nodes on the pipeline, calculates the correlation peak and time delay between the signals, performs secondary location of the leak point, and obtains a second location result.

[0024] A further proposed approach is to include a model reverse verification step in the subsequent verification process:

[0025] The dynamic hydraulic model simulation unit is triggered. Taking the real-time collected flow and pressure data as input, it performs inverse hydraulic verification and leakage scenario simulation through iterative fitting, and outputs a virtual leak point location.

[0026] Calculate the second Euclidean distance between the virtual leak point location and the geometric centers of the first and second positioning results;

[0027] If the second Euclidean distance is less than the second preset distance tolerance, the hydraulic model verification is deemed successful, and the positioning information with the virtual leak point location as the final leak point location is output.

[0028] If the second Euclidean distance is greater than or equal to the second preset distance tolerance, the verification is deemed to have failed, and only an alarm prompt is output without outputting precise positioning information.

[0029] A further proposed approach is to trigger the hydraulic condition for reverse verification of the model as follows:

[0030] Within any partition, the following coordination anomaly conditions must be met:

[0031] The measured values ​​of the zone inlet flow sensors are consistently higher than the theoretical inlet flow calculated by the dynamic hydraulic model simulation unit under the condition of no leakage, and the positive deviation exceeds the first flow threshold.

[0032] Meanwhile, the measured value of the pressure sensor at at least one node in the partition is consistently lower than the theoretical node pressure value calculated by the dynamic hydraulic model simulation unit under the condition of no leakage, and the negative deviation exceeds the first pressure threshold.

[0033] In a further embodiment, the specific steps of the iterative fitting include:

[0034] The leakage event is parameterized as the pipe number of the zone where the leak point is located, the relative location of the pipe, and the amount of water leakage;

[0035] A loss function is constructed to quantify the difference between the partition and node hydraulic simulation data and the real-time acquired data under the virtual leak point parameters;

[0036] An optimization algorithm is used to iteratively solve the problem with the goal of minimizing the loss function, outputting the optimal virtual leak parameters and their corresponding simulated confidence scores, which are then input into the fusion analysis module.

[0037] A further solution is that the loss function Specifically:

[0038] ;

[0039] in, This represents the nodes within the partition obtained through hydraulic simulation calculations under the set virtual leak point parameters. Pressure value;

[0040] This indicates the nodes within the partition as measured in real time by a pressure sensor. Pressure value;

[0041] This represents the inlet flow rate of the partition obtained through hydraulic simulation calculation under the set virtual leak point parameters;

[0042] This represents the inlet flow rate of the zone, measured in real time by a flow sensor.

[0043] and These are preset weighting coefficients;

[0044] Σ represents the summation calculation of all nodes within the partition that have pressure monitoring capabilities;

[0045] This indicates the number of nodes within the partition that have pressure monitoring capabilities.

[0046] A further embodiment of the proposed tiered early warning information includes:

[0047] When the first Euclidean distance is greater than or equal to the first preset distance tolerance and the model reverse verification step is not triggered, a first type of warning information is output, which includes an abnormal status prompt.

[0048] When the first Euclidean distance is less than the first preset distance tolerance, a second type of warning information is output. The second type of warning information includes missing location information with the geometric center of the first location result and the second location result as the coordinates.

[0049] When the second Euclidean distance is less than the second preset distance tolerance, a third type of early warning information is output, which includes leakage location information verified by the hydraulic model.

[0050] In a further embodiment, the negative pressure wave localization algorithm and the acoustic correlation analysis algorithm are mutually adjusted during execution, specifically as follows:

[0051] The fusion analysis module further includes the following steps: first, running the negative pressure wave positioning algorithm to output a first positioning interval; then, inputting the boundary parameters of the first positioning interval into the acoustic correlation analysis algorithm to constrain its signal correlation analysis search range, and outputting the second positioning result.

[0052] In a further embodiment, the negative pressure wave localization algorithm and the acoustic correlation analysis algorithm are mutually adjusted iteratively, and the fusion analysis module performs the following steps:

[0053] S1. Run the negative pressure wave localization algorithm and the acoustic correlation analysis algorithm in parallel to obtain the initial first localization result and the second localization result;

[0054] S2. Calculate the third Euclidean distance between the initial first positioning result and the second positioning result;

[0055] S3. If the third Euclidean distance is greater than a preset iteration tolerance, the adjustment process is triggered: the midpoint of the line connecting the two positioning results is taken as the new search reference point.

[0056] S4. Using the new search reference point as the center, within a reduced search radius, re-run the negative pressure wave positioning algorithm and / or acoustic correlation analysis algorithm to obtain updated first positioning results and / or second positioning results;

[0057] S5. Repeat steps S1 to S4 until the third Euclidean distance is less than or equal to the iteration tolerance, or the preset maximum number of iterations is reached.

[0058] The beneficial effects of this invention are:

[0059] (1) This invention abandons the drawbacks of traditional single-algorithm dependence and susceptibility to interference. It innovatively calls the negative pressure wave positioning algorithm based on pressure transients and the acoustic correlation analysis algorithm based on pipeline noise in parallel, and introduces Euclidean distance as a consistency verification criterion, thereby ensuring the credibility of the positioning results: when the positioning results obtained by the two algorithms with different physical principles are highly consistent, the system can confidently output accurate positioning; when the results are significantly different, a higher-level verification process is triggered, effectively avoiding false alarms. Compared with a single algorithm, the dual verification mechanism greatly enhances the adaptability to complex pipeline network environments such as background noise and pressure fluctuations, solves the industry pain points of inaccurate positioning and low reliability of existing technologies, and provides solid and reliable data support for subsequent maintenance decisions.

[0060] (2) When the positioning results of the pressure and acoustic sensor algorithm are inconsistent, the present invention performs reverse verification by triggering a dynamic hydraulic model. The dynamic hydraulic model takes the real-time collected flow and pressure data as input, simulates the leakage scenario through iterative fitting, and outputs an independent virtual leak location. This virtual location will be compared with the geometric center of the sensor positioning result in a second distance, thus forming a rigorous verification step. When the simulation results of the hydraulic model and the sensor positioning can be mutually verified, the system outputs the verified high-confidence positioning. When they cannot be verified, the system chooses to only alarm and not output the potentially incorrect positioning, thereby reducing the risk of false alarms and avoiding ineffective excavation and resource waste caused by incorrect positioning. Attached Figure Description

[0061] The invention will now be further described with reference to the accompanying drawings.

[0062] Figure 1 This is a flowchart of an IoT-based regional water supply network leakage analysis and early warning system. Detailed Implementation

[0063] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0064] Please see Figure 1 As shown, this invention is an IoT-based regional water supply network leakage analysis and early warning system, comprising:

[0065] The zoning module is used to divide the water supply network into multiple independent zones and to deploy IoT sensor arrays at the inlet, outlet and internal key nodes of each zone; the key nodes are pipe intersections, elevation change points and monitoring points at preset distance intervals.

[0066] The fusion analysis module runs on a cloud platform and is communicatively connected to the sensor array.

[0067] The fusion analysis module is used to execute the verification process, which outputs the leakage location result through the following steps:

[0068] The negative pressure wave localization algorithm based on pressure transients and the acoustic correlation analysis algorithm based on pipeline noise are invoked in parallel to generate the first localization result and the second localization result, respectively.

[0069] Calculate the first Euclidean distance between the first positioning result and the second positioning result;

[0070] Based on the comparison result between the first Euclidean distance and the first preset distance tolerance, determine whether to output the final location information of the leak point or trigger the subsequent verification process.

[0071] The early warning module receives the output from the cloud platform analysis module, generates tiered early warning information, and displays it on a visualization map based on partitions and nodes. Before the fusion analysis module performs Euclidean distance calculation and loss function optimization, all pressure, flow, and location coordinate data are first preprocessed to eliminate the influence of dimensions and improve the convergence and accuracy of the algorithm.

[0072] The sensors include a pressure sensor, a flow meter, and a noise sensor;

[0073] The comparison process between the first Euclidean distance and the first preset distance tolerance is as follows:

[0074] If the first Euclidean distance is less than the first preset distance tolerance, it is determined that the positioning is consistent, and the positioning information with the geometric center of the two positioning results as the final leak point position is output; for example, combining the first positioning result and the second positioning result, the midpoint of the line connecting them is taken as the final leak point position.

[0075] If the first Euclidean distance is greater than or equal to the first preset distance tolerance, it is determined that the positioning is inconsistent and the subsequent verification process is triggered.

[0076] A further solution is that the negative pressure wave localization algorithm captures the instantaneous negative pressure wave generated when a pipeline leaks, and uses the time difference between the time the pressure sensors deployed at both ends of the pipeline receive the negative pressure wave to perform preliminary localization of the leak point and obtain a first localization result.

[0077] The principle of the negative pressure wave localization algorithm is that when a pipeline experiences a sudden leak, a negative pressure wave with an instantaneous pressure drop will be generated at the leak point due to fluid leakage. This negative pressure wave propagates to both ends of the pipeline at a certain speed of sound.

[0078] The specific positioning steps include:

[0079] By using high-frequency pressure sensors installed at both ends of the pipeline (node ​​A and node B) to capture transient pressure signals and filtering and noise reduction of the signals, the origin of the negative pressure wave can be accurately identified.

[0080] Accurately calculate the time difference between the arrival of the negative pressure wave at sensor A and sensor B. .

[0081] Leakage location calculation: based on the formula Calculate the distance between the leak point and sensor A.

[0082] in:

[0083] This is the distance from the leak point to sensor A.

[0084] This is the total length of the pipe (i.e., the distance between sensors A and B).

[0085] The speed of sound is the propagation speed of a negative pressure wave in the fluid of a pipe. This speed can be estimated by the pipe material, pipe diameter and fluid properties, or calibrated by field test.

[0086] The first positioning result can be obtained by following the above steps.

[0087] The acoustic correlation analysis algorithm analyzes the leakage noise signals received by two noise sensors deployed at two different nodes on the pipeline, calculates the correlation peak and time delay between the signals, performs secondary location of the leak point, and obtains a second location result.

[0088] The acoustic correlation analysis algorithm works on the principle that continuous noise generated by leakage propagates through the pipe wall and fluid to distant locations, where it is received by noise sensors placed at different locations in the pipeline. Due to the different distances between the sound source (leak point) and the two sensors, there is a time delay in the noise signal reaching the two sensors.

[0089] The specific positioning steps include:

[0090] Two noise sensors synchronously collect noise signals over a period of time.

[0091] The cross-correlation function of the two acquired signals is calculated. The x-axis corresponding to the peak value of the cross-correlation function (time delay) is determined. This is the time difference between the arrival of the noise signal at the two sensors.

[0092] Leakage location calculation: based on the formula Calculate the distance between the leak point and the first sensor.

[0093] in:

[0094] This represents the distance from the leak point to the first sensor.

[0095] The length of the pipe between the two noise sensors.

[0096] The speed of sound is the propagation velocity of leakage noise in the pipe. This speed is related to the pipe structure and can be determined in advance through experiments.

[0097] The second positioning result can be obtained by following the above steps.

[0098] In this solution, the pipeline network is divided into independent zones through a zoning module, and pressure, flow, and noise sensor arrays are deployed at key nodes (inlet, outlet, intersection, elevation change points), thereby achieving refined pipeline network management and comprehensive monitoring. Decomposing the large system into small units enables rapid location of leakage anomalies to specific zones, greatly narrowing the scope of investigation and laying the foundation for rapid response.

[0099] Based on this, the negative pressure wave localization algorithm (utilizing the time difference of pressure wave transmission) and the acoustic correlation analysis algorithm (utilizing the correlation of leakage noise signals) are run in parallel on the cloud platform. This utilizes the complementarity of different physical principles and overcomes the limitations of a single algorithm. The negative pressure wave method responds quickly to sudden leakage, while the acoustic method is sensitive to continuous leakage. The parallel operation of the two ensures that the system can effectively capture and provide preliminary localization for any type of leakage.

[0100] Calculate the first Euclidean distance between the positioning results of the two algorithms and compare it with the first preset distance tolerance. If the distance is less than the tolerance, it means that the conclusions of the two independent methods are consistent, and the system can confidently output accurate positioning (based on the geometric center), which greatly reduces the risk of false alarms. If the distance is greater than or equal to the tolerance, it means that there is a discrepancy in the positioning results. The system does not output unreliable positioning, but triggers an abnormal alarm, prompting manual intervention or initiating subsequent verification processes, thus avoiding invalid repairs caused by erroneous information.

[0101] The above scheme has several advantages. First, the zoned management significantly reduces the scope of the loss investigation and improves the positioning efficiency. Second, the two positioning algorithms based on different physical principles effectively complement each other, overcoming the limitations of a single algorithm. Most importantly, the consistency verification mechanism of Euclidean distance establishes a reliable decision-making basis. When the conclusions of the two independent methods are consistent, the system can output accurate positioning, greatly reducing the risk of false alarms. When the results are inconsistent, the system avoids outputting unreliable positioning, preventing invalid maintenance caused by erroneous information.

[0102] For example:

[0103] Water supply branch pipes in the public area of ​​the waiting hall of a high-speed railway station.

[0104] The water supply branch pipes in this area are divided into the "West Waiting Hall Water Supply-01" zone. A flow meter is installed at the inlet of this branch pipe, a pressure sensor is installed at the beginning and end of the pipe (50 meters apart), and a noise sensor is installed in the middle section of the pipe (approximately 25 meters away).

[0105] A pipe joint located approximately 28 meters from the starting point is experiencing a continuous leak due to loosening.

[0106] Parallel analysis and localization:

[0107] Negative pressure wave location algorithm: When a leak occurs, a weak pressure transient is generated, which is captured by pressure sensors at both ends of the pipeline. By analyzing the time difference between the two pressure signals, the system calculates the initial location of the leak, L1 = 26 meters (distance from the starting end).

[0108] Acoustic correlation analysis algorithm: The hissing noise generated by the leakage is collected by a noise sensor. The system separates and performs correlation analysis on this signal with the background noise from the pipeline fluid and the environment, and calculates the location of the leak, L2 = 29 meters.

[0109] Consistency verification and decision-making:

[0110] The system calculates the first Euclidean distance between L1 and L2: |29-26|=3 meters.

[0111] The system calls the preset first distance tolerance for this partition. This tolerance is set to 10% of the total pipe length of 50 meters, or 5 meters.

[0112] Judgment: 3 meters < 5 meters. The system determines that the two positioning results are consistent, and the verification passes.

[0113] Output result:

[0114] The system automatically generates a second type of early warning information.

[0115] The final positioning result is the geometric center of L1 and L2: (26+29) / 2=27.5 meters.

[0116] The location information (“West Waiting Hall Water Supply-01” zone, 27.5 meters from the starting point) was sent to the operation and maintenance management platform along with the alarm and accurately marked on the visual map. Maintenance personnel can then directly access the target area for precise repairs based on this information, avoiding large-scale excavation and lengthy investigations.

[0117] The subsequent verification process includes a model reverse verification step:

[0118] The dynamic hydraulic model simulation unit is triggered. Taking the real-time collected flow and pressure data as input, it performs inverse hydraulic verification and leakage scenario simulation through iterative fitting, and outputs a virtual leak point location.

[0119] Calculate the second Euclidean distance between the virtual leak point location and the geometric centers of the first and second positioning results;

[0120] If the second Euclidean distance is less than the second preset distance tolerance, the hydraulic model is deemed to have passed verification, and the location information with the virtual leak location as the final leak location is output. The first preset distance tolerance and the second preset distance tolerance can be set to be proportional to the average length of the pipes in the partition. For example, the first preset distance tolerance can be set to 10% of the average pipe length (i.e., 10 meters of pipe, tolerance of 10 meters); the second preset distance tolerance can be set to 20% of the average pipe length (i.e., 100 meters of pipe, tolerance of 20 meters).

[0121] If the second Euclidean distance is greater than or equal to the second preset distance tolerance, the verification is deemed to have failed, and only an alarm prompt is output without outputting precise positioning information.

[0122] The hydraulic condition that triggers the reverse verification of the model is:

[0123] Within any partition, the following coordination anomaly conditions must be met:

[0124] The measured values ​​of the zone inlet flow sensors are consistently higher than the theoretical inlet flow calculated by the dynamic hydraulic model simulation unit under the condition of no leakage, and the positive deviation exceeds the first flow threshold.

[0125] Meanwhile, the measured value of the pressure sensor at at least one node within this partition is consistently lower than the theoretical node pressure value calculated by the dynamic hydraulic model simulation unit under leak-free conditions, and the negative deviation exceeds the first pressure threshold. "Consistently" means that the abnormal state remains stable for more than a preset time window, such as 5 minutes, to eliminate instantaneous fluctuation interference.

[0126] The first flow threshold and the first pressure threshold can be set based on the statistical fluctuation range of the monitoring data during the historical normal operation of the partition. For example, the first flow threshold can be set to twice the standard deviation of the historical flow data at the partition entrance; the first pressure threshold can be set to twice the standard deviation of the historical pressure data of the node.

[0127] The specific steps of the iterative fitting include:

[0128] The leakage event is parameterized as the pipe number of the zone where the leak point is located, the relative location of the pipe, and the amount of water leakage;

[0129] A loss function is constructed to quantify the difference between the partition and node hydraulic simulation data and the real-time acquired data under the virtual leak point parameters;

[0130] An optimization algorithm is employed to iteratively solve for the loss function, outputting the optimal virtual leak parameters and their corresponding simulated confidence scores, which are then input into the fusion analysis module. The optimization algorithm can be either a particle swarm optimization algorithm or a genetic algorithm.

[0131] The loss function Specifically:

[0132] ;

[0133] in, This represents the nodes within the partition obtained through hydraulic simulation calculations under the set virtual leak point parameters. Pressure value;

[0134] This indicates the nodes within the partition as measured in real time by a pressure sensor. Pressure value;

[0135] This represents the inlet flow rate of the partition obtained through hydraulic simulation calculation under the set virtual leak point parameters;

[0136] This represents the inlet flow rate of the zone, measured in real time by a flow sensor.

[0137] and The weighting coefficients are preset; and This is used to balance the contribution weights of pressure and flow observations in the loss function, and its specific value can be configured based on the relative reliability and magnitude differences of pressure and flow monitoring data.

[0138] For example, in a partition with dense pressure monitoring nodes and high data reliability, it is possible to set up =0.7, =0.3, to focus more on pressure fitting; while in a partition with extremely high inlet flow meter accuracy and sparse pressure nodes, it can be set to 0.3. =0.4, =0.6, to be more dependent on flow fit; the weighting coefficients satisfy + =1 is the normalization relation.

[0139] Σ represents the summation calculation of all nodes within the partition that have pressure monitoring capabilities;

[0140] This indicates the number of nodes within the partition that have pressure monitoring capabilities.

[0141] The tiered early warning information includes:

[0142] When the first Euclidean distance is greater than or equal to the first preset distance tolerance and the model reverse verification step is not triggered, a first type of warning information is output, which includes an abnormal status prompt.

[0143] When the first Euclidean distance is less than the first preset distance tolerance, a second type of warning information is output. The second type of warning information includes missing location information with the geometric center of the first location result and the second location result as the coordinates.

[0144] When the second Euclidean distance is less than the second preset distance tolerance, a third type of early warning information is output, which includes leakage location information verified by the hydraulic model.

[0145] This solution constructs a condition-triggered, multi-stage verification intelligent decision-making closed loop, which is triggered when the preliminary positioning results obtained through two algorithms—pressure transient and acoustic correlation analysis—are inconsistent. Its workflow begins with the determination of the triggering conditions: the system continuously monitors the measured data from the inlet flow sensors and node pressure sensors of each zone and compares them with the theoretical values ​​calculated by the dynamic hydraulic model under leakage-free conditions. Only when the positive flow deviation exceeds a first flow threshold set based on the standard deviation of historical data, and the negative pressure deviation exceeds the corresponding first pressure threshold, and this abnormal state remains stable for more than a preset time window (e.g., 5 minutes) to eliminate instantaneous interference, will the system determine that the coordinated abnormal conditions are met, thus initiating the model reverse verification process. This rigorous triggering mechanism ensures efficient utilization of system resources and avoids unnecessary computational overhead.

[0146] After entering the model reverse verification phase, the dynamic hydraulic model simulation unit begins operation, parameterizing the leakage event into three key variables: the pipe number where the leak point is located, the relative position of the pipe, and the leakage volume. The system quantifies the difference between the simulated data and the real-time acquired data under the virtual leak point parameters by constructing a loss function. This function is typically in the form of a weighted sum of squares, balancing the contributions of pressure and flow observations. Subsequently, intelligent algorithms such as particle swarm optimization or genetic algorithms are used to iteratively solve the problem with the goal of minimizing the loss function. This process essentially involves repeatedly simulating the leakage scenario in a digital twin environment, ultimately outputting the optimal virtual leak point location and its simulation confidence level.

[0147] Next, the system performs a second verification: it calculates the second Euclidean distance between the virtual leak location and the geometric center of the previous sensor positioning result, and compares this distance with a more lenient second preset distance tolerance.

[0148] If the distance is less than the tolerance limit, the model validation is successful, and the system outputs a third-type warning containing precise positioning information validated by the hydraulic model. If the distance exceeds the tolerance limit, the system cautiously outputs only an abnormal status alert without providing positioning information, thus effectively avoiding misleading maintenance decisions. The entire process reflects a progressive analysis approach from data perception to model validation, ensuring high reliability and confidence of the output results through multiple verification mechanisms.

[0149] The above technical solution addresses two key issues. First, when sensor algorithms based on different physical principles yield conflicting location results, the system no longer simply chooses one or abandons the location. Instead, it utilizes dynamic hydraulic model simulation—a verification channel independent of the sensor network—using real-time collected flow and pressure data as boundary conditions. Through iterative calculation, it inversely derives the most probable leak location, providing a final basis for maintenance decisions and reducing the risk of misjudgment due to algorithm limitations or environmental interference. Second, it achieves precise grading of early warning information confidence. When the model inversion result matches the sensor location area, it outputs high-confidence location information that has undergone dual verification. This dual guarantee mechanism of preliminary sensor location + final model confirmation allows maintenance departments to confidently conduct precise excavation based on system guidance, greatly improving emergency response efficiency. Conversely, when the model and sensor conclusions still cannot be reconciled, the system wisely issues only an anomaly alarm without providing location information. This prudent strategy effectively avoids ineffective excavation, resource waste, and secondary damage to the pipeline network caused by providing incorrect locations.

[0150] For example:

[0151] Inside the underground utility tunnel of a high-speed railway station, a key water supply main pipe, about 100 meters long, is responsible for supplying water to the entire west waiting hall and commercial area.

[0152] Initial state:

[0153] Zoning and Model: The main pipe is divided into the "West Main Pipe-01" zone. A dynamic hydraulic model of this pipe section has been pre-built in the system, which includes all physical parameters such as pipe friction, elevation, and valve status.

[0154] Threshold settings: Based on the historical normal operation data of this partition, the system sets the following: the first flow threshold is twice the historical flow standard deviation (i.e., 0.8 L / s), the first pressure threshold is twice the historical pressure standard deviation (i.e., 1.5 meters head); the second preset distance tolerance is set to 20% of the pipe length, i.e., 20 meters.

[0155] The sequence of events:

[0156] One day, the system detected that the measured value of the inlet flow sensor in this zone was 1.2 L / s higher than the theoretical value for more than 5 minutes, exceeding the threshold of 0.8 L / s.

[0157] Meanwhile, the measured value of the pressure sensor at a key node in the utility tunnel was consistently 2.0 meters lower than the theoretical value, exceeding the threshold of 1.5 meters.

[0158] The system determines that the collaborative anomaly conditions are met and automatically triggers the model reverse verification process.

[0159] Model back-validation and iterative fitting:

[0160] The dynamic hydraulic model simulation unit is started, using the current abnormal inlet flow and the measured pressure values ​​of all nodes as the fitting targets.

[0161] The iterative process begins: the optimization algorithm (such as particle swarm optimization) begins to try thousands of sets of virtual leak point parameters.

[0162] First, assuming the leak is 50 meters away and the leakage rate is 5 L / s, we run a hydraulic simulation and find that the calculated pressure is far from the measured value (i.e., the loss function L is very large).

[0163] Next, we tried a 10-meter depth with a leakage rate of 10 L / s, but the simulation results still did not match.

[0164] After multiple iterations, the algorithm found that when the parameters were set at 75 meters from the starting point and the leakage rate was 8.5 L / s, the simulated pressure values ​​and inlet flow rates at each point were most consistent with the actual measured data on site (the loss function L was minimized).

[0165] Secondary verification and decision output:

[0166] The system outputs the virtual leak location L3 = 75 meters.

[0167] Because the positioning results from the pressure transient method and the acoustic method were too different (for example, one reported 30 meters and the other reported 50 meters) and failed the initial verification, the system calculated the second Euclidean distance between L3 (75 meters) and their geometric center (40 meters) to be 35 meters.

[0168] Decision judgment: 35 meters > 20 meters (second preset distance tolerance). The system determined that the model verification failed.

[0169] Final warning:

[0170] The system eventually generated and output the first type of early warning information, which reads: "A hydraulic anomaly has been confirmed in the West Main Pipeline-01 section, suspected to be a serious leak, but the exact location cannot be provided at present. Please send personnel to inspect immediately."

[0171] Upon receiving this alarm, although the exact location is unknown, the maintenance personnel have identified the fault zone and its severity. They can immediately organize technicians to enter the pipe gallery with portable testing equipment to conduct a focused inspection of the 100-meter pipe section, thereby quickly finding the actual leak.

[0172] The negative pressure wave localization algorithm and the acoustic correlation analysis algorithm are mutually adjusted during execution, specifically as follows:

[0173] The fusion analysis module further includes the following steps: first, running the negative pressure wave positioning algorithm to output a first positioning interval; then, inputting the boundary parameters of the first positioning interval into the acoustic correlation analysis algorithm to constrain its signal correlation analysis search range, and outputting the second positioning result.

[0174] This invention is a serially guided algorithm adjustment strategy; its specific implementation lies in changing the execution logic of the negative pressure wave positioning algorithm and the acoustic correlation analysis algorithm, changing it from completely parallel processing to sequential collaborative work.

[0175] The fusion analysis module first runs the negative pressure wave positioning algorithm. This algorithm utilizes the high propagation speed of pressure waves to quickly respond to leakage events. However, its positioning accuracy is greatly affected by factors such as wave speed estimation errors. Therefore, the system does not use its output as a precise positioning point, but rather as a preliminary, range-based first positioning interval.

[0176] Subsequently, the system uses the boundary parameters of this interval as key inputs to the acoustic correlation analysis algorithm. Although the acoustic algorithm involves a large amount of computation and is sensitive to environmental noise, its local positioning accuracy can theoretically be very high. By constraining the search range of the acoustic algorithm to the interval given by the negative pressure wave algorithm, the system effectively shields it from interference from a large number of invalid signals outside the interval, allowing it to concentrate its computational resources on fine-grained signal correlation analysis and time delay calculation within the suspected area, thereby outputting a more accurate and reliable second positioning result.

[0177] This scheme achieves a balanced optimization of positioning efficiency and accuracy, fully leveraging the advantages of both algorithms to form a highly efficient working mode of coarse positioning guiding fine positioning. The negative pressure wave algorithm acts as a rapid screening tool, quickly narrowing down the search area; the acoustic analysis algorithm plays the role of precise measurement, achieving accurate positioning within the narrowed area. This serial guidance mechanism brings two direct benefits: First, it significantly reduces unnecessary computational load on the acoustic algorithm, improving the overall response speed of the system, especially suitable for sudden large-loss scenarios requiring rapid response; second, by limiting the search range, it effectively reduces the probability of the acoustic algorithm being interfered with by noise outside the interval, thereby improving the reliability and accuracy of the final positioning result.

[0178] The negative pressure wave localization algorithm and the acoustic correlation analysis algorithm are adjusted to each other iteratively. The fusion analysis module performs the following steps:

[0179] S1. Run the negative pressure wave localization algorithm and the acoustic correlation analysis algorithm in parallel to obtain the initial first localization result and the second localization result;

[0180] S2. Calculate the third Euclidean distance between the initial first positioning result and the second positioning result;

[0181] S3. If the third Euclidean distance is greater than a preset iteration tolerance, the adjustment process is triggered: the midpoint of the line connecting the two positioning results is taken as the new search reference point.

[0182] S4. Using the new search reference point as the center, within a reduced search radius, re-run the negative pressure wave positioning algorithm and / or acoustic correlation analysis algorithm to obtain updated first positioning results and / or second positioning results;

[0183] S5. Repeat steps S1 to S4 until the third Euclidean distance is less than or equal to the iteration tolerance, or the preset maximum number of iterations is reached. The iteration tolerance can be set as a proportion of the first preset distance tolerance, for example, 50%, and the maximum number of iterations can be set to a fixed value, for example, 5 to 10, based on the required balance between computational accuracy and response speed.

[0184] In this scheme, the negative pressure wave localization algorithm and the acoustic correlation analysis algorithm are first run in parallel to obtain their respective independent initial localization results, namely the first localization result and the second localization result. The system then calculates the third Euclidean distance between these two results and compares it with a preset iteration tolerance, which is usually more stringent than the final localization accuracy requirement. If the distance is greater than the tolerance, it indicates that the initial conclusions of the two algorithms are significantly different and have low reliability. At this point, the system does not simply take the average output, but triggers an iterative adjustment process: the midpoint of the line connecting the two localization results is taken as a new, more reliable search reference point. Subsequently, the system re-runs one or both algorithms with this midpoint as the center within a significantly reduced search radius. This process can be repeated, with each iteration further narrowing and optimizing the search range around a more reliable reference point, forcing the output results of the two algorithms to converge within a continuously shrinking high-quality solution space, until the third Euclidean distance between them is less than or equal to the iteration tolerance, or the maximum number of iterations is reached.

[0185] The technical advantage of this scheme lies in achieving autonomous optimization and convergence of positioning results, greatly enhancing the system's robustness and positioning accuracy in complex interference environments. By establishing a closed-loop feedback mechanism, when the initial positioning is unreliable, the system can automatically initiate multiple rounds of iterative optimization, using information provided by the two algorithms to mutually correct and guide each other, ultimately converging to a consensus solution recognized by both. This mechanism is particularly suitable for scenarios with complex acoustic environments and variable pipeline conditions, effectively addressing situations where a single algorithm malfunctions due to accidental interference. Even if one algorithm deviates due to sudden noise in a calculation, the other algorithm will bring it back on track in the next iteration.

[0186] It should be noted that the calculation formulas and all parameters involved in the calculations in this invention have been dimensionless beforehand. The process of dimensionless processing is well known in the industry and will not be described here.

[0187] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. An Internet of Things-based regional water supply network leakage analysis and early warning system, characterized in that, The system comprises: a partition module for partitioning a water supply network into multiple independent partitions and arranging an Internet of Things sensor array for each partition inlet, outlet and internal key node; a fusion analysis module running on a cloud platform and in communication connection with the sensor array; the fusion analysis module is configured to execute a verification process to output a leakage location result by the following steps: parallelly calling a negative pressure wave location algorithm based on pressure transient and an acoustic correlation analysis algorithm based on pipe noise to generate a first location result and a second location result respectively; calculating a first Euclidean distance between the first location result and the second location result; judging to output location information of a final leakage point position or triggering a subsequent verification process according to a comparison result of the first Euclidean distance and a first preset distance tolerance; an early warning module for receiving an output of the cloud platform analysis module, generating hierarchical early warning information and displaying on a visual map based on partitions and nodes; the sensors include pressure sensors, flow meters and noise sensors; the comparison process of the first Euclidean distance and the first preset distance tolerance is as follows: if the first Euclidean distance is less than the first preset distance tolerance, it is determined that the location is consistent, and location information of a geometric center of the two location results as the final leakage point position is outputted; if the first Euclidean distance is greater than or equal to the first preset distance tolerance, it is determined that the location is inconsistent, and a subsequent verification process is triggered; the negative pressure wave location algorithm captures the transient negative pressure wave generated when leakage occurs in the pipe, and uses the time difference of the negative pressure wave received by the pressure sensors arranged at the two end nodes of the pipe to preliminarily locate the leakage point and obtain the first location result; the acoustic correlation analysis algorithm analyzes the leakage noise signals received by the two noise sensors arranged at two different nodes of the pipe, calculates the correlation peak value and time delay between the signals, and performs secondary location of the leakage point to obtain the second location result; the subsequent verification process includes a model reverse verification step: triggering a dynamic hydraulic model simulation unit to input real-time flow and pressure data, performing reverse hydraulic checking and leakage scenario simulation through iterative fitting, and outputting a virtual leakage point position; calculating a second Euclidean distance between the virtual leakage point position and the geometric center of the first and second location results; if the second Euclidean distance is less than a second preset distance tolerance, it is determined that the hydraulic model verification is passed, and location information of the virtual leakage point position as the final leakage point position is outputted; if the second Euclidean distance is greater than or equal to the second preset distance tolerance, it is determined that the verification is not passed, and only an alarm prompt is outputted without outputting accurate location information. 2.The IoT-based regional water supply network leakage analysis and early warning system according to claim 1, characterized in that, the hydraulic condition for triggering the model reverse verification is as follows: in any partition, the following cooperative anomaly condition is met: the measured value of the partition inlet flow sensor is continuously higher than the inlet flow theoretical value calculated by the dynamic hydraulic model simulation unit under the no-leakage working condition, and the positive deviation amplitude exceeds a first flow threshold value; Meanwhile, the measured value of the pressure sensor of at least one node in the partition continuously falls below the theoretical value of the node pressure calculated by the dynamic hydraulic model simulation unit under the no-leakage working condition, and the negative deviation amplitude exceeds a first pressure threshold. 3.The IoT-based regional water supply network leakage analysis and early warning system according to claim 1, characterized in that, The specific steps of the iterative fitting include: Parameterizing the leakage event as the pipe number, pipe relative position, and water leakage amount of the partition where the leakage point is located; Constructing a loss function for quantifying the difference between the hydraulic simulation data of the partition under the virtual leakage point parameters and the real-time collected data; Using an optimization algorithm to iteratively solve the loss function with the goal of minimizing the loss function, output the optimal virtual leakage point parameters and their corresponding simulation confidence, and input them to the fusion analysis module. 4.The IOT-based regional water supply network leakage analysis and early warning system according to claim 3, characterized in that, The loss function Specifically: ； wherein, represents the pressure value of the node in the subarea calculated by hydraulic simulation under the set virtual leak point parameter; represents the pressure value of the node in the subarea calculated by hydraulic simulation under the set virtual leak point parameter; represents a pressure value of the node in the partition measured in real time by the pressure sensor; represents a pressure value of the node in the partition measured in real time by the pressure sensor; represents the partition inlet flow value calculated by hydraulic simulation under the set virtual leak point parameter; represents the zone inlet flow value measured in real time by the flow sensor; and are preset weight coefficients; Σ represents the summation calculation of all nodes with pressure monitoring capability in the partition; To indicate the number of nodes in the partition that have pressure monitoring capability. 5.The IoT-based regional water supply network leakage analysis and early warning system according to claim 1, characterized in that, The hierarchical early warning information includes: When the first Euclidean distance is greater than or equal to the first preset distance tolerance and the model reverse verification step is not triggered, output the first type of early warning information, which contains an abnormal state prompt; When the first Euclidean distance is less than the first preset distance tolerance, output the second type of early warning information, which contains leakage location information with the geometric center of the first positioning result and the second positioning result as the coordinates; When the second Euclidean distance is less than the second preset distance tolerance, output the third type of early warning information, which contains leakage location information verified by the hydraulic model. 6.The IoT-based regional water supply network leakage analysis and early warning system according to claim 1, characterized in that, The negative pressure wave positioning algorithm and the acoustic correlation analysis algorithm are adjusted to each other when they are executed, specifically as follows: The fusion analysis module further includes the following steps: first, run the negative pressure wave positioning algorithm to output a first positioning interval; then, input the boundary parameters of the first positioning interval to the acoustic correlation analysis algorithm to constrain the search range of its signal correlation analysis, and output the second positioning result. 7.The IoT-based regional water supply network leakage analysis and early warning system according to claim 1, characterized in that, The negative pressure wave positioning algorithm and the acoustic correlation analysis algorithm are adjusted to each other through an iterative manner, and the fusion analysis module executes the following steps: S1, run the negative pressure wave positioning algorithm and the acoustic correlation analysis algorithm in parallel to obtain initial first positioning results and second positioning results; S2, calculate the third Euclidean distance between the initial first positioning results and the second positioning results; S3, if the third Euclidean distance is greater than a preset iteration tolerance, trigger the adjustment process: take the midpoint of the line connecting the two positioning results as a new search reference point; S4, take the new search reference point as the center, and re-run the negative pressure wave positioning algorithm and / or the acoustic correlation analysis algorithm within a reduced search radius to obtain updated first positioning results and / or second positioning results; S5, repeat steps S1 to S4 until the third Euclidean distance is less than or equal to the iteration tolerance, or the maximum number of iterations is reached.

Images