A water supply network leakage detection and positioning method and system

Abstract

This invention relates to the field of smart water management technology, providing a method and system for detecting and locating leaks in water supply networks. It collects multi-dimensional data on network operation in real time by deploying pressure sensors, flow meters, and acoustic sensors at key nodes of the water supply network. Simultaneously, it combines static and dynamic information such as network GIS data, water usage patterns, and meteorological data. Secondly, it constructs a long short-term memory neural network model based on an attention mechanism to perform deep feature extraction and fusion of the collected multi-source time-series data, learning the complex dynamic patterns of the network under normal and leaking conditions. Finally, the model performs real-time analysis of the network status. When an abnormal pattern is detected, the system automatically triggers an alarm and uses backpropagation and graph theory algorithms to accurately calculate the possible location of the leak point. This invention significantly improves the accuracy, timeliness, and location precision of leak detection, effectively reducing the operating costs and water waste for water supply companies.

Description

Technical Field

[0001] This invention relates to the field of smart water management technology, and in particular to a method and system for detecting and locating leaks in water supply networks. Background Technology

[0002] Urban water supply networks are the "lifeline" for maintaining the normal operation of cities, but leakage has always been a serious challenge for the global water industry. Network leakage not only causes a huge waste of precious water resources and increases the production and operating costs of water supply companies, but can also trigger secondary disasters such as ground subsidence and water pollution, threatening public safety.

[0003] For example, application number CN202110707699.5 discloses a method and system for detecting and locating leaks in a water supply network. The method includes: acquiring the location of pressure sensors in the water supply network and collecting water supply pressure data for a period of time; establishing a bidirectional long short-term memory network model and training the bidirectional long short-term memory network; calculating the channel distance between each monitoring point and all pressure monitoring points to obtain a distance matrix, and performing negative and normalization processing to obtain a distance weight matrix; obtaining the pressure prediction value of the monitoring point based on the long short-term memory network, calculating the residual between the actual value and the predicted value of each monitoring point to obtain a residual matrix, normalizing the residual matrix, and constructing an error matrix based on the distance weight; finding abnormal points through box plots, and judging the leakage of the water supply network based on a pre-set leakage judgment time threshold and an abnormal monitoring point number threshold.

[0004] Existing technologies generally suffer from low detection accuracy and poor real-time performance, making it impossible to achieve all-weather, high-precision, automated, and intelligent water supply network leakage detection. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides a method and system for detecting and locating leaks in water supply networks.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for detecting and locating leaks in a water supply network includes the following steps: S1: Real-time collection of dynamic operation data through a sensor network deployed in the water supply network, and acquisition of static and external data related to the operation of the network to form a multi-source heterogeneous dataset; S2: Construct a deep learning-based leakage identification model, and train the model using historical multi-source datasets to enable it to learn the data feature patterns under normal and leakage conditions of the pipeline network. S3: Input the real-time collected multi-source data into the trained leakage identification model, and the model outputs the leakage probability value of each area of ​​the current pipeline network; S4: When the probability value of leakage in a certain area exceeds the preset threshold, the system determines that leakage has occurred in that area, and starts the positioning algorithm. Combining the pipeline topology, it calculates and outputs the most likely location coordinates of the leakage point, and generates alarm information at the same time.

[0007] Preferably, the dynamic operation data in step S1 includes: pressure time-series data of key nodes in the pipeline network, flow time-series data of pipe sections, and pipeline noise signal spectrum data collected by acoustic sensors; the static and external data include: pipeline geographic information system data, pipe material age data, user water usage pattern data, and meteorological data.

[0008] Preferably, the leakage identification model in step S2 is a long short-term memory neural network based on the attention mechanism. This model first processes time-series data from different sources through multiple LSTM units, then uses the attention mechanism to weight and fuse the features output by different LSTM units, and finally outputs the leakage probability through a fully connected layer.

[0009] Preferably, the training process of the model includes: using historical data containing normal operating conditions and confirmed leakage conditions as training samples, using whether leakage occurs and its location as labels, optimizing model parameters through backpropagation algorithm, and minimizing the loss function between the prediction result and the true label.

[0010] Preferably, the localization algorithm in step S4 specifically involves: using the area with the highest leakage probability as the center, and combining the direction of pressure gradient change and the acoustic signal intensity attenuation model, applying Dijkstra's algorithm or... on the pipeline topology map. The algorithm searches for the best matching path between pressure anomalies and acoustic anomalies, and the intersection of the paths is the predicted leakage point.

[0011] A water supply network leakage detection and location system includes: a data acquisition module, a data preprocessing and storage module, an intelligent analysis module, a leakage location and alarm module, and a human-computer interaction and visualization module. The data acquisition module collects pressure, flow, and acoustic data in real time through a sensor network deployed on the pipeline network and connects to external data sources such as GIS, water usage patterns, and meteorological data. The data preprocessing and storage module cleans, aligns, and normalizes the collected multi-source heterogeneous data and stores the processed data in a time-series database. The intelligent analysis module has a built-in leakage identification model based on an attention mechanism-LSTM, used to receive real-time data, analyze the pipeline network's operating status, and calculate the leakage probability of each area. The leakage location and alarm module, when the detected leakage probability exceeds a threshold, activates a location algorithm to calculate the location of the leakage point and generates a visualized alarm containing information such as location and severity, which is then pushed to maintenance personnel. The human-computer interaction and visualization module provides a graphical interface to display the real-time operating status of the pipeline network, leakage alarm information, historical data analysis reports, and supports the configuration and optimization of model parameters.

[0012] Preferably, the sensor network in the data acquisition module uses low-power wide-area network technology for data transmission to achieve large-scale, low-cost deployment.

[0013] Preferably, the intelligent analysis module is deployed on a cloud server or edge computing node to utilize its powerful computing capabilities for real-time inference of deep learning models.

[0014] The advantages of this invention are as follows: by integrating multi-dimensional information such as pressure, flow rate, and acoustics, and utilizing AI models to deeply mine the complex correlations between data, it can detect minute, early leaks that are difficult to detect using traditional methods. Furthermore, by combining hydraulic characteristics, acoustic characteristics, and pipeline topology for comprehensive positioning, the positioning accuracy can reach the meter level, significantly reducing the scope and time of manual inspection. The system operates 24 / 7 without interruption, and the entire process from data acquisition to alarm generation is completed automatically, achieving a leak response within seconds or minutes. This reduces a large amount of manual inspection and leak listening work, lowering labor costs. Moreover, rapid leak location can reduce water loss and repair excavation costs, improving overall operational efficiency. Attached Figure Description

[0015] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a simplified flowchart of the present invention; Figure 2 This is a flowchart of the present invention; Figure 3 This is a system architecture diagram of the present invention; Detailed Implementation

[0016] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0017] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0018] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0019] Example 1, combined with Figures 1-3 Explanation: In the city's water supply network, 500 key nodes were selected and equipped with high-precision pressure sensors, electromagnetic flow meters, and acoustic sensors with filtering functions. The sensors transmit data to the data center once per hour or at a frequency set as needed via NB-IoT technology. At the same time, the system is connected to the city's water supply GIS database, SCADA system database, user billing system database, and the API interface of the National Meteorological Administration. The data fusion module aligns all data to the minute level according to timestamps and associates sensor data with pipe section and node information based on GIS coordinates.

[0020] By installing high-precision pressure sensors, it's possible to quickly detect whether the hydraulic balance has been disrupted. Electromagnetic flow meters can provide a macroscopic assessment of whether a leak has occurred in a given area. Acoustic sensors can monitor the noise generated at the leak point by the high-pressure water jet rubbing against the pipe wall and surrounding medium. By linking these three sensors, a comprehensive monitoring network is formed: pressure changes provide early warnings, flow rate changes provide macroscopic location information, and acoustic detection provides precise localization, thus efficiently locating leaks.

[0021] By using NB-IoT technology to send data at scheduled frequencies as needed, the technology offers advantages such as low power consumption, wide coverage, and low cost, thus addressing the feasibility of large-scale deployment.

[0022] By connecting to external GIS, SCADA, billing, and meteorological data sources, the system can make more intelligent and accurate judgments based on external data information.

[0023] The water supply GIS database provides the topology and pipe segment attributes of the pipeline network. The system can determine different leakage probabilities based on different conditions. For example, the leakage probability is different when an abnormal pressure occurs on a cast iron pipe and when the same abnormality occurs on a PE pipe. The normal pressure fluctuation range of an old pipeline network may be larger than that of a new pipeline network.

[0024] SCADA systems monitor the operational status of water plants, pumping stations, and large valves. When SCADA displays that a pumping station is starting or stopping, the pipeline pressure will change drastically. If SCADA data is not connected, this normal scheduling may be misjudged as a large-scale pipe burst. After connecting to the SCADA system, it will be treated as a planned operation, thus avoiding false alarms.

[0025] The user billing system database can provide historical and real-time water usage patterns for each region and each type of user (residential, commercial, industrial), thus effectively distinguishing between normal water usage and abnormal leakage in the pipeline network.

[0026] The National Meteorological Administration's API interface can provide external environmental data, such as the possibility that a sudden drop in temperature may cause pipes to contract and lead to abnormal pressure; during heavy rain, surface water may seep into the pipe network, affecting water quality and pressure; and during hot weather, residents' water consumption will increase significantly. By analyzing external environmental data, interference from environmental factors can be eliminated, thereby improving the robustness of the detection.

[0027] By employing a data fusion module, data containing all information such as pressure, flow rate, sound, temperature, and rainfall can be unified at a given point in time, thereby enabling the correlation of numerous data points.

[0028] GIS coordinate association can give spatial meaning to data, and can correspond abstract sensor readings to specific physical entities. This allows the system to mark abnormal locations on the map, and then combine them with pipe segment attributes for analysis, and finally perform precise positioning.

[0029] In summary, through cross-validation of multi-source data and contextual analysis, the system can effectively distinguish between real leakage and normal scheduling, user water usage, environmental interference, etc., reducing the false alarm rate to an extremely low level.

[0030] The input is multi-source time-series data (pressure, flow rate, acoustic spectrum main frequency, etc.) from the past 24 hours. By using multi-source time-series data from the past 24 hours as input, a complete time window can be provided for the system. Only by comparing the changing trends at the same time yesterday and the previous few hours can an anomaly be determined.

[0031] Three parallel LSTM layers are set up to process these three types of data respectively. The three LSTM layers are pressure LSTM layer, flow LSTM layer, and acoustic LSTM layer. The goal of the pressure LSTM layer is to know what kind of pressure fluctuations are normal pump station start-ups and shutdowns, and what kind of fluctuations are suspicious small drops. The goal of the flow LSTM layer is to be able to distinguish between peak residential water consumption and unexplained continuous flow increases. The goal of the acoustic LSTM layer is to be able to identify specific hissing sounds that represent water leakage from noisy background noise.

[0032] The attention layer calculates the weights of the output of each LSTM unit. For example, when there are abnormal fluctuations in pressure, the output weights of the pressure LSTM will automatically increase. The fused feature vector passes through two fully connected layers and finally outputs a leakage probability. It is trained using historical data from the past three years, which includes 120 confirmed leakage events and their detailed records. After 200 epochs of training, the model achieves an accuracy of 98% and a recall of 95% on the test set.

[0033] After the system went live, real-time data streams were fed into the trained model. The model scanned the entire pipeline network every 5 minutes and output a leakage probability score for each pipe segment. This setup, through high-frequency scanning, transformed the complex pipeline network status into a quantified risk score in real time, thereby enabling accurate early warning and proactive management of leaks.

[0034] At 3:00 AM one day, the system detected that the leakage probability score in area A rapidly increased from 0.1 to 0.92, exceeding the threshold of 0.85. The system immediately determined that a leakage had occurred in the area, and the location algorithm was activated: analyzing the data from eight pressure sensors in and around area A, it was found that the lowest pressure point was located between sensors P3 and P4; at the same time, acoustic sensor M3 detected an abnormal noise signal with a much higher intensity than other sensors. Combining this with the GIS map, the system determined that the leakage was most likely to occur on a section of DN300 cast iron pipe connecting P3 and P4 and close to M3, and provided the specific GIS coordinates. The alarm information was immediately sent to the mobile app of the night shift emergency repair team.

[0035] The system architecture of this embodiment includes: Data acquisition module: Composed of on-site sensors, data acquisition terminals, and a communication network, it is used to acquire raw data. Sensors directly contact the pipeline network, converting physical quantities such as pressure, flow rate, and sound into electrical signals. The data acquisition terminal connects to multiple sensors, responsible for collecting and initially processing the data (such as simple filtering and packaging), and periodically or on-demand transmitting the data through the communication network. The communication network utilizes technologies such as NB-IoT and 4G to ensure reliable and low-cost transmission of data from underground pipelines throughout the city to the cloud data center.

[0036] Data preprocessing and storage module: Deployed in the cloud, using Kafka for data stream processing, InfluxDB for time-series data storage, and PostgreSQL for GIS and business data. Kafka can receive and temporarily store all data at extremely high speeds, preventing data surges from overwhelming subsequent systems. InfluxDB is a time-series database; the biggest characteristic of sensor data is that it contains timestamps and is write-heavy with few reads. InfluxDB is extremely optimized for this data type, with extremely fast write speeds, small storage footprint, and efficient querying by time range. PostgreSQL is a relational database with powerful spatial data extensions (PostGIS), making it ideal for storing data that changes infrequently but has a complex structure and strong correlations. GIS data includes pipe segment coordinates, lengths, materials, and connection relationships; business data includes the correspondence between sensor IDs and pipe segments, user information, maintenance records, etc.

[0037] Intelligent Analysis Module: Deployed on a cloud server with GPU, it uses the TensorFlow or PyTorch framework to implement AI models and provides API interfaces for other modules to call.

[0038] The Leakage Location and Alert Module is a microservice that can subscribe to the results of the intelligent analysis module, execute location logic, and push alerts to users via a message queue.

[0039] Human-computer interaction and visualization module: This is a web-based front-end application that operations and maintenance personnel can access through a browser to view real-time data and alarm flashing points on the pipeline GIS map, and can drill down to view historical curves and model analysis reports.

[0040] A highly efficient and scalable system architecture was built through modular design, realizing closed-loop management of the entire process from data collection to intelligent decision-making and then to visualized operation and maintenance.

[0041] The working principle of this invention is as follows: A sensor network deployed in the water supply network collects dynamic operational data in real time and acquires static and external data related to network operation, forming a multi-source heterogeneous dataset. A deep learning-based leakage detection model is constructed and trained using historical multi-source datasets, allowing it to learn data feature patterns under normal and leaking network conditions. The real-time collected multi-source data is input into the trained leakage detection model, which outputs the leakage probability value for each area of ​​the current network. When the leakage probability value of a certain area exceeds a preset threshold, the system determines that a leakage has occurred in that area and activates a location algorithm. Combining the network topology, the system calculates and outputs the most likely location coordinates of the leakage point, while simultaneously generating an alarm message. This invention significantly improves the accuracy, timeliness, and location precision of leakage detection, effectively reducing the operating costs of water supply companies and water resource waste.

[0042] For those skilled in the art, the present invention is not limited to the details of the exemplary embodiments described above, and can be implemented in other specific forms without departing from the spirit or essential characteristics of the invention; therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Therefore, it is intended that all variations falling within the meaning and scope of equivalents of the claims be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0043] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any minor modifications, equivalent substitutions, and improvements made to the above embodiments based on the technical essence of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for detecting and locating leaks in a water supply network, characterized in that, Includes the following steps: S1: Real-time collection of dynamic operation data through a sensor network deployed in the water supply network, and acquisition of static and external data related to the operation of the network to form a multi-source heterogeneous dataset; S2: Construct a deep learning-based leakage identification model, and train the model using historical multi-source datasets to enable it to learn the data feature patterns under normal and leakage conditions of the pipeline network. S3: Input the real-time collected multi-source data into the trained leakage identification model, and the model outputs the leakage probability value of each area of ​​the current pipeline network; S4: When the probability value of leakage in a certain area exceeds the preset threshold, the system determines that leakage has occurred in that area, and starts the positioning algorithm. Combining the pipeline topology, it calculates and outputs the most likely location coordinates of the leakage point, and generates alarm information at the same time.

2. The method for detecting and locating leakage in a water supply network according to claim 1, characterized in that, The dynamic operation data in step S1 includes: pressure time-series data of key nodes in the pipeline network, flow time-series data of pipe sections, and pipeline noise signal spectrum data collected by acoustic sensors; the static and external data include pipeline geographic information system data, pipe material age data, user water usage pattern data, and meteorological data.

3. The method for detecting and locating leakage in a water supply network according to claim 1, characterized in that, The leakage identification model in step S2 is based on a long short-term memory neural network with an attention mechanism. This model processes time-series data from different sources through multiple LSTM units, then uses the attention mechanism to weight and fuse the features output by different LSTM units, and finally outputs the leakage probability through a fully connected layer.

4. The method for detecting and locating leakage in a water supply network according to claim 3, characterized in that, The training process of the model includes: using historical data containing normal operating conditions and confirmed leakage conditions as training samples, using whether leakage occurs and its location as labels, optimizing model parameters through backpropagation algorithm, and minimizing the loss function between the predicted results and the true labels.

5. The method for detecting and locating leakage in a water supply network according to claim 1, characterized in that, The localization algorithm in step S4 specifically involves: centering on the area with the highest leakage probability, and combining the direction of pressure gradient change and the acoustic signal intensity attenuation model, using Dijkstra's algorithm or... on the pipeline topology map. The algorithm searches for the best matching path between pressure anomalies and acoustic anomalies, and the intersection of multiple anomaly paths is the predicted leakage location.

6. A water supply network leakage detection and location system, characterized in that, The system is used to implement the method of any one of claims 1 to 5. The system includes: a data acquisition module, a data preprocessing and storage module, an intelligent analysis module, a leakage location and alarm module, and a human-computer interaction and visualization module. The data acquisition module is used to collect pressure, flow, and acoustic data in real time through a sensor network deployed on the pipeline network, and to access external data sources such as GIS, water usage patterns, and meteorological data. The data preprocessing and storage module is used to clean, align, and normalize the collected multi-source heterogeneous data, and store the processed data in a time-series database. The intelligent analysis module has a built-in leakage identification model based on the attention mechanism-LSTM, used to receive real-time data, analyze the pipeline network operation status, and calculate the leakage probability of each area. The leakage location and alarm module is used to activate a location algorithm to calculate the location of the leakage point when the leakage probability exceeds a threshold, and generate a visual alarm containing information such as location and severity, which is then pushed to maintenance personnel. The human-computer interaction and visualization module provides a graphical interface to display the real-time operation status of the pipeline network, leakage alarm information, historical data analysis reports, and supports the configuration and optimization of model parameters.

7. A water supply network leakage detection and location system according to claim 6, characterized in that, The sensor network in the data acquisition module uses low-power wide-area network technology for data transmission to achieve large-scale, low-cost deployment.

8. A water supply network leakage detection and location system according to claim 6, characterized in that, The intelligent analysis module is deployed on a cloud server or edge computing node to leverage its powerful computing capabilities for real-time inference of deep learning models.

Images