A method and system for dynamic analysis of DMA partition leakage of a water supply network and a medium

Abstract

The application relates to the technical field of water supply network management, and discloses a water supply network DMA partition leakage dynamic analysis method, a system and a medium, the method comprising the following steps: acquiring pipeline data and real-time monitoring data in a DMA partition of a water supply network, obtaining a partition physical characteristic index and an operation state vector, based on the partition physical characteristic index and the operation state vector, adopting a classification algorithm to extract a potential leakage signal from the real-time monitoring data, and combining a weighting matrix and dynamic path analysis to determine a candidate position of a leakage point; accurately positioning the leakage point and generating a propagation path atlas through neighborhood connection data and matching degree calculation; and calculating the propagation weight of each path according to the propagation path, and optimizing the positioning of the leakage point and the repair priority. The application can significantly improve the efficiency of leakage detection and reduce false positives and false negatives.

Description

Technical Field

[0001] This application relates to the field of water supply network management technology, and in particular to a dynamic analysis method, system and medium for DMA zone leakage in water supply networks, specifically applied to leakage detection and location in urban water supply networks. Background Technology

[0002] With the continuous advancement of urbanization, the management and maintenance of water supply systems face increasingly complex challenges. As a crucial component of urban infrastructure, the operational efficiency and water resource utilization efficiency of water supply networks directly impact the sustainable development of cities. However, in actual operation, leakage frequently occurs in water supply networks. This leakage not only increases water waste but also places enormous pressure on the stability and economy of urban water supply systems.

[0003] Traditional methods for detecting leaks in water supply networks often rely on single data sources and static analysis models, which are ineffective in addressing the complex dynamic changes within the network. Especially in large-scale urban water supply networks, accurate location and timely detection of leaks remain a critical challenge. Existing methods typically suffer from the following problems: low accuracy in identifying leak signals and a high false alarm rate; difficulty in tracing the propagation path of leak events and accurately determining the location of the leak source; and a lack of effective dynamic analysis tools, failing to reflect abnormal states in network operation in real time.

[0004] DMA (Districted Management Area) is a water supply network management model that improves the flexibility and response speed of water supply network management by dividing the network into multiple independent zones for monitoring. However, how to utilize real-time monitoring data based on DMA zoning, combined with multi-dimensional data such as pipeline physical characteristics, pressure fluctuations, and flow changes, to accurately and dynamically analyze and locate leakage events remains a major technological challenge.

[0005] Therefore, in response to the need for dynamic analysis of leakage in water supply networks, this application provides a dynamic analysis method for leakage in water supply networks by DMA partitioning. The method aims to improve the accuracy and efficiency of leakage detection, reduce the false alarm rate, and trace the propagation path of leakage events through the fusion of multi-dimensional data and dynamic correlation analysis, thereby achieving accurate location and real-time management of leakage in water supply networks. Summary of the Invention

[0006] This application provides a method, system, and medium for dynamic analysis of leakage in water supply networks via DMA (Distributed Area Mapping) zones, which can improve the detection accuracy and location efficiency of leakage events and optimize the operation and maintenance management of water supply networks.

[0007] In a first aspect, this application provides a method for dynamic analysis of leakage in a water supply network via DMA zones, the method comprising:

[0008] S1. Obtain pipeline data and real-time monitoring data within the DMA partition of the water supply network to obtain the partition's physical characteristic indicators and operating status vector;

[0009] S2. Based on the partition physical characteristic indicators and the running state vector, a classification algorithm is used to determine the sequence of potential leakage events;

[0010] S3. Extract time series features from the potential leakage event sequence, and combine them with a preset threshold to determine the preliminary leakage signal, and generate a preliminary leakage signal set;

[0011] S4. For the initial set of leakage signals, fuse pipeline data to construct a weighted matrix and generate a weighted pressure fluctuation matrix;

[0012] S5. Through the correlation analysis between the weighted pressure fluctuation matrix and the time series features, a feature extraction algorithm is used to determine the list of candidate locations for leakage points;

[0013] S6. Obtain neighborhood connection data from the candidate location list of the leakage point, and determine the high probability leakage point by combining the matching degree to obtain the precise location coordinate set;

[0014] S7. Generate a dynamic correlation map based on the precise positioning coordinate set to obtain the propagation path of the leakage event;

[0015] S8. Optimize the path weights of the dynamic association graph through optimization algorithms to determine the final leakage points.

[0016] Secondly, this application provides a dynamic analysis system for leakage in a water supply network DMA zone, the system comprising:

[0017] The data collection unit is used to acquire pipeline data and real-time monitoring data within the DMA partition of the water supply network, and to obtain the partition's physical characteristic indicators and operating status vector.

[0018] The event determination unit is used to determine a sequence of potential leakage events based on the partition physical characteristic indicators and the running state vector using a classification algorithm;

[0019] The feature extraction unit is used to extract time series features from the potential leakage event sequence, and combine them with a preset threshold to determine the preliminary leakage signal and generate a preliminary leakage signal set.

[0020] The matrix construction unit is used to construct a weighted matrix by fusing pipeline data with the preliminary leakage signal set, and generate a weighted pressure fluctuation matrix.

[0021] The correlation analysis unit is used to determine a list of candidate locations for leakage points by using a feature extraction algorithm through correlation analysis between the weighted pressure fluctuation matrix and the time series features.

[0022] The location matching unit is used to obtain neighborhood connection data from the list of candidate locations of the leak points, and combine the matching degree to determine the high probability of leak points to obtain a precise location coordinate set;

[0023] The map generation unit is used to generate a dynamic correlation map based on the precise positioning coordinate set to obtain the propagation path of the leakage event.

[0024] The path optimization unit is used to optimize the path weights of the dynamic association graph through optimization algorithms to determine the final leakage points.

[0025] A third aspect of this application provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the aforementioned method for dynamic analysis of leakage in a DMA partition of a water supply network.

[0026] Compared with the prior art, the beneficial effects of the present invention are at least as follows:

[0027] The technical solution provided in this application firstly generates physical characteristic indicators and operating status vectors for the water supply network by acquiring pipeline data and real-time monitoring data within the DMA (Digital Transmission Area) partition. Then, using these physical characteristic indicators and operating status vectors, a classification algorithm is employed to determine potential leakage event sequences. Next, preliminary leakage signals are judged by extracting time-series features and combining them with preset thresholds, generating a preliminary leakage signal set. Subsequently, a weighted matrix is ​​constructed and a weighted pressure fluctuation matrix is ​​generated. Further, through correlation analysis with time-series features, a list of candidate leakage point locations is determined. Next, neighborhood connection data is acquired and high-probability leakage points are calculated using matching degrees, resulting in a precise location coordinate set. A dynamic correlation map is generated based on the precise location coordinate set, and finally, the map path weights are optimized using an optimization algorithm to determine the final leakage point location. This application can achieve precise leakage detection and dynamic monitoring in complex water supply networks, improving the response speed and accuracy of leakage detection, reducing false alarm rates, and increasing the efficiency of leakage detection. Attached Figure Description

[0028] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a schematic diagram of an embodiment of a dynamic analysis method for leakage in a water supply network DMA zone according to the present application.

[0030] Figure 2This is a schematic diagram of an embodiment of a dynamic analysis system for leakage in a water supply network DMA partitioning according to the present application. Detailed Implementation

[0031] This application provides a method, system, and medium for dynamic analysis of leakage in a water supply network using a DMA (Distributed Area Mitigation) partition. The terms "first," "second," "third," "fourth," etc. (if present)," in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0032] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the dynamic analysis method for leakage in a water supply network DMA zone in this application includes:

[0033] S1. Obtain pipeline data and real-time monitoring data within the DMA partition of the water supply network to obtain the partition's physical characteristic indicators and operating status vector.

[0034] S2. Based on the partition physical characteristic indicators and the operating status vector, a classification algorithm is used to determine the sequence of potential leakage events.

[0035] S3. Extract time series features from the potential leakage event sequence, and combine them with a preset threshold to determine the initial leakage signal, and generate an initial leakage signal set.

[0036] S4. For the initial set of leakage signals, integrate pipeline data to construct a weighted matrix and generate a weighted pressure fluctuation matrix.

[0037] S5. Through correlation analysis between the weighted pressure fluctuation matrix and time series features, a feature extraction algorithm is used to determine the list of candidate locations for leakage points.

[0038] S6. Obtain neighborhood connection data from the candidate location list of leakage points, combine it with the matching degree to determine the high probability of leakage points, and obtain the precise location coordinate set.

[0039] S7. Generate a dynamic correlation map based on the precise positioning coordinate set to obtain the propagation path of leakage events.

[0040] S8. Optimize the path weights of the dynamic association graph through optimization algorithms to determine the final leakage points.

[0041] By coordinating the above steps, this application can effectively improve the detection accuracy and location efficiency of leakage events.

[0042] In one specific embodiment, the process of performing the above steps further includes:

[0043] S1. Acquire pipeline data and real-time monitoring data within the DMA (Digital Distributed Area) zone of the water supply network to obtain the zone's physical characteristic indicators and operating status vector. Specifically, this includes collecting diameter distribution ratio data, real-time pressure data, and flow monitoring data of pipelines with different diameters within the DMA zone of the water supply network; fusing time-series pressure data with pipeline diameter ratio data; calculating pipe diameter weights; adjusting real-time pressure data and flow monitoring data based on pipe diameter weights; and generating zone physical characteristic indicators. Combining the zone physical characteristic indicators with real-time monitoring data, an operating status vector is constructed, which characterizes the dynamic operating features of the DMA zone of the water supply network.

[0044] Specifically, by deploying sensor devices at key nodes of the pipeline network DMA partition, real-time data on the diameter distribution ratio of pipes of different diameters, i.e., pipe diameter ratio data, is collected, including the proportion of small, medium, and large pipe diameters. At the same time, real-time pressure data from pressure sensors and instantaneous readings from flow meters are recorded, i.e., flow monitoring data. Real-time pressure data and flow monitoring data constitute real-time monitoring data. The continuously collected real-time pressure data is arranged in chronological order to form time-series pressure data, which is represented as a sequence vector, where each element corresponds to the pressure value at a specific time point. Based on the pipe diameter ratio data, a scaling factor is applied to the pressure sequence of different pipe diameters to calculate the pipe diameter weight.

[0045] The specific process for calculating pipe diameter weights includes: normalizing the time-series pressure data, mapping the pressure values ​​to the range of 0 to 1 to eliminate dimensional differences, defining a weighting function based on the pipe diameter ratio data (e.g., assigning higher weights to large-diameter pipes with higher ratios), and fusing the data using a weighted average formula to obtain the overall pipe diameter weight value. For example, in a city water supply network DMA zone, where small-diameter pipes account for 40%, medium-diameter pipes 30%, and large-diameter pipes 30%, and the time-series pressure data sequence is [2.5, 2.6, 2.4] bar, when fusing and calculating the weights, the data is first normalized to obtain [0.5, 0.52, 0.48], and then weighted according to the ratio to obtain a weight of 0.65. This method is beneficial for capturing the structural differences in pressure changes within the pipe zone, avoiding bias caused by single data points, and improving the accuracy of subsequent adjustments to real-time pressure and flow monitoring data.

[0046] By using the calculated pipe diameter weights, real-time pressure and flow data are weighted and corrected. For example, pressure data is multiplied by a weighting factor, and flow data is similarly adjusted. This process generates comprehensive indicators, such as average pressure deviation or flow stability values, to quantify the physical characteristics of the zones. During the adjustment process, if the pipe diameter weight is 0.65, the real-time pressure is adjusted from 2.5 bar to 1.625 bar, and the flow rate is adjusted from 100 L / s to 65 L / s. Then, indicators such as pressure fluctuation variance are calculated to be 0.1. This method helps improve the accuracy of indicators in representing the actual state of the pipeline network and facilitates early identification of leakage.

[0047] By combining zoned physical characteristic indicators with real-time monitoring data, an operational status vector is constructed. Specifically, this involves concatenating physical characteristic indicators, such as pressure deviation, with real-time data, such as flow rate, to form a multi-dimensional vector for further analysis of the pipeline network's operational status. Furthermore, by extracting key features from real-time monitoring data, such as peak flow and average pressure, and concatenating them with physical characteristic indicators, the resulting operational status vector effectively captures the dynamic changes in the pipeline network, supporting subsequent anomaly detection and leakage identification. This method solves the problem of existing methods failing to effectively reflect the dynamic characteristics of the pipeline network and enables analysis of the pipeline network based on its actual operational status. Real-time monitoring of the network improves the efficiency of leakage event identification. In specific implementation, for example, in a residential area's pipe network section, the physical characteristic index is [2.3, 0.08], and the real-time data is [90, 2.4]. By splicing them together to form a vector [2.3, 0.08, 90, 2.4], this vector can effectively analyze the state changes during low flow at night, which helps improve the efficiency of leakage event identification. In addition, under seasonal water use changes, this method can accurately reflect the dynamic characteristics of the pipe network by analyzing the changes in flow components, thereby supporting more accurate pipe network management and leakage detection.

[0048] S2. Based on the partition physical characteristic indicators and the running state vector, a classification algorithm is used to determine the sequence of potential leakage events. Specifically, the partition physical characteristic indicators and the running state vector are input into the support vector machine algorithm, and the abnormal pattern is classified by the support vector machine algorithm to obtain the classification result. Based on the classification result, the fluctuation amplitude of the running state vector is calculated. If the fluctuation amplitude exceeds the preset fluctuation threshold, the corresponding running state vector is determined as a potential leakage event and a sequence of potential leakage events is generated. Otherwise, the corresponding running state vector is determined as a non-potential leakage event.

[0049] Specifically, this method collects ratio data of pipes with different diameters and real-time pressure and flow monitoring values ​​within the old pipeline network zones. Combined with time-series pressure data and pipe diameter weight adjustments, it generates zone physical characteristic indicators and operational status vectors. The zone physical characteristic indicators reflect the structural features of the pipeline network, such as pipe diameter ratios and pipe length distribution, while the operational status vectors contain real-time monitored pressure and flow data. By inputting the zone physical characteristic indicators and operational status vectors into a support vector machine (SVM) algorithm, a hyperplane is constructed in a high-dimensional space to distinguish between normal and abnormal patterns. Furthermore, the radial basis function kernel of the SVM algorithm is used to process nonlinear data, mapping the zone physical characteristic indicators and operational status vectors to a high-dimensional space, thereby identifying abnormal patterns of leakage events such as abnormal pressure drops or sudden flow increases, and obtaining classification results. This method solves the problem that traditional methods cannot effectively identify leakage events in complex pipeline networks, significantly improving the accuracy of leakage detection and reducing the false alarm rate.

[0050] After classifying and obtaining abnormal operating state vectors, the fluctuation amplitude of these vectors is calculated. Specifically, statistical analysis is performed on the pressure sequence of each vector, and its standard deviation is calculated to quantify the severity of the anomaly. If the calculated fluctuation amplitude exceeds a preset fluctuation threshold, the vector is marked as a potential leakage event, forming a potential leakage event sequence. For example, if a vector contains pressure values ​​at 10 time points, such as 50, 48, 52, 45, 55, 42, 58, 40, 60, and 38, and its standard deviation is approximately 7.8, exceeding the preset threshold of 5.0, the vector will be considered a potential leakage event. This method can effectively quantify the degree of pressure fluctuation anomalies in the pipeline network and determine potential leakage events based on the fluctuation amplitude, ensuring the continuity of events and thus providing data support for analyzing leakage propagation paths. In older pipe networks, especially in pipes with larger diameters, pressure changes are more likely to trigger large-scale leaks. Therefore, by adjusting the weights and emphasizing the fluctuation range of large-diameter pipes, potential leak points can be identified more accurately. For example, if a vector shows a pressure drop from the normal 50 to 38, with an amplitude of 8.2, exceeding the threshold, it is identified as a potential leak event, and its timestamp is recorded for easy tracking of the event's evolution. In contrast, for partitions with balanced pipe diameter ratios, the calculated fluctuation range is 4.5, which does not exceed the threshold of 5.0, and therefore is not included in the potential leak event sequence, avoiding false positives and optimizing resource allocation. This method, through support vector machine classification combined with historical data training, further improves its adaptability to real-time monitoring scenarios, enabling timely identification and intervention of leak events and effectively improving the accuracy of leak detection.

[0051] S3. Extract time series features from the potential leakage event sequence and combine them with a preset threshold to determine the preliminary leakage signal and generate a preliminary leakage signal set. Specifically, this includes extracting time series features from the potential leakage event sequence, where the time series features include the flow rate change rate and pressure fluctuation frequency. Based on the time series features, determine whether the flow anomaly exceeds the preset flow threshold. If the flow anomaly exceeds the preset flow threshold, combine the anomaly pattern classification results to determine the preliminary leakage signal, summarize all preliminary leakage signals, and generate a preliminary leakage signal set.

[0052] Specifically, time series features, including flow rate change rate and pressure fluctuation frequency, are extracted from potential leakage event sequences. The flow rate change rate is obtained by dividing the difference in flow rate between adjacent time points by the time interval. For example, in zoned pipe network monitoring, if the sequence shows that the flow rate drops from 100 liters / min to 80 liters / min from t1 to t2, the change rate is -20 liters / min per unit time. The pressure fluctuation frequency is obtained by analyzing the periodicity of the pressure data through Fourier transform to obtain the number of fluctuations per unit time. Time series features help capture sudden changes in the pipe network and support the accuracy of anomaly detection. By comparing these features with preset flow thresholds, when the flow abnormally exceeds the preset threshold, a preliminary leakage signal is determined by combining the abnormal pattern classification results. The support vector machine algorithm, in this process, constructs a hyperplane in a high-dimensional space to distinguish between normal and abnormal patterns, thereby improving the recognition accuracy of the leakage signal. This method can better adapt to complex pipeline structures, improve the fault tolerance of leakage detection, reduce false positives, and optimize pipeline maintenance efficiency. For example, in old pipeline networks, when the flow abnormally exceeds the threshold in a large-diameter pipe in a certain section, and the classification result is a high-probability leakage pattern, combining these two factors generates a preliminary leakage signal, thereby improving the efficiency of leakage event location. By summarizing all determined preliminary leakage signals and organizing them into a set in chronological order, complete data is formed, facilitating the tracking and analysis of leakage events. This method can effectively improve the accuracy of leakage detection and reduce the false alarm rate.

[0053] S4. For the initial leakage signal set, a weighted matrix is ​​constructed by fusing pipeline data to generate a weighted pressure fluctuation matrix. Specifically, this includes acquiring pressure fluctuation data from the initial leakage signal set, fusing pipe diameter ratio data, adjusting the weights of the pressure fluctuation data to generate weighted pressure values, constructing a weighted matrix based on the correlation between the weighted pressure values ​​and flow monitoring data, and calculating the coupling relationship between pressure fluctuations and flow anomalies through the weighted matrix to generate the weighted pressure fluctuation matrix.

[0054] Specifically, pressure change values ​​related to the time series are extracted from the preliminary leakage signal set by real-time monitoring equipment to reflect the fluctuation characteristics of potential anomalies and help identify abnormalities in the pipeline network. After acquiring the pressure fluctuation data, preliminary filtering is performed to remove noise. To reduce random interference, a moving average method is used to smooth the pressure series. Taking smoothing with a window size of 5 as an example, the original pressure series [10.2, 10.5, 9.8, 10.0, 10.3] is smoothed to obtain [10.2, 10.166, 10.1, 10.033, 10.2]. This eliminates fluctuations in the data, effectively solves the noise interference problem, improves the stability of pressure fluctuation data, provides accurate data support for weight adjustment and leakage event location, and optimizes the leakage detection process of the water supply network.

[0055] By integrating pipe diameter ratio data and adjusting the weights of pressure fluctuation data, a weighted pressure value is generated to address the impact of pipeline network structure differences on leakage detection. In practice, the pipe diameter ratio data originates from the proportion of different pipe diameters within a pipeline network section, such as 0.6 for large diameters, 0.3 for medium diameters, and 0.1 for small diameters. These ratios are used as weighting factors and multiplied by the corresponding pipe's pressure fluctuation data to obtain a weighted pressure value. For example, for a large diameter pressure fluctuation value of 10.0, applying a weighting factor of 0.6 yields a weighted pressure value of 6.0, which better reflects the impact of large diameter pipes on overall fluctuations. By calculating the weighting factor for each pipe diameter category and normalizing the pipe diameter ratio data, the total weight is ensured to be 1, thus avoiding adjustment bias. The weighting factors are applied point-by-point to multiply the pressure fluctuation data to generate a preliminary weighted sequence. For example, for a pressure fluctuation value P_t = 10.0 at a certain time point and a pipe diameter weight W_i = 0.6, the weighted sequence is calculated as follows: The weighted pressure value A_t is 6.0; subsequently, the exponential moving average method is used to smooth the initial weighted sequence to capture long-term trends and further improve the stability of leakage detection; for example, if the proportion of large-diameter pipes in the pipeline network is 0.7, medium-diameter pipes 0.2, and small-diameter pipes 0.1, in the pressure fluctuation sequence [9.5, 9.8, 9.2], the weighted value generated by applying the weights for the large-diameter portion is [6.65, 6.86, 6.44]. This adjustment emphasizes the sensitivity of large-diameter pipes. This method helps identify hidden leaks. During peak flow periods, the pipe diameter ratio is dynamically adjusted, with the weight of larger pipe diameters increasing to 0.65. When applied to the pressure sequence [11.0, 10.7, 11.2], the weighted pressure values ​​[7.15, 6.955, 7.28] are obtained, which can effectively improve the accuracy of response to flow fluctuations. This method can improve the accuracy of leak detection and ensure the reliability of leak analysis, especially in the face of complex pipe network structures and dynamic flow changes, thus optimizing the leak monitoring of the pipe network.

[0056] This study analyzes the correlation between weighted pressure values ​​and flow monitoring data by constructing a weighted matrix, addressing the inaccurate identification of leakage events in existing technologies. In implementation, the weighted pressure value sequence is paired with concurrent flow monitoring data, and the correlation coefficient between the two continuous variables is calculated as a measure of correlation. For example, if the correlation coefficient between the pressure sequence [6.0, 6.3] and the flow data [50, 52] is 0.95, it indicates a strong positive correlation between pressure and flow. A matrix is ​​constructed using the correlation coefficient, where each row represents a time point and each column represents the correlation value of the pressure-flow pair. Specifically, the flow monitoring data is aligned with the weighted pressure value sequence by matching timestamps; for example, pressure value t1=6.0 corresponds to flow 50, and t2=6.3 corresponds to flow 52, ​​forming a paired dataset. Then, based on the paired data, the correlation coefficient formula is used to calculate the correlation coefficient for each... The correlation coefficient is calculated as r = cov(pressure, flow) / (std(pressure) * std(flow)). For example, if cov = 0.15, std(pressure) = 0.15, and std(flow) = 1.0, then r = 1.0, indicating a strong positive correlation. By filling the above correlation indicators into the matrix, a weighted matrix is ​​formed. By constructing the weighted matrix, the multidimensional correlation between pressure and flow is fully captured, improving the ability to identify abnormal patterns in the pipeline network. In specific implementation, if the correlation coefficient calculated between the weighted pressure value sequence [6.0, 6.3, 5.9] and the flow data [50, 52, 48] is 0.98, then the constructed matrix is ​​{[0.98, 0.97], [0.97, 0.98]}, which reflects the close linkage between pressure and flow, helping to detect abnormal patterns early and optimize the detection of leakage events.

[0057] By calculating the coupling relationship between pressure fluctuations and flow anomalies, a weighted pressure fluctuation matrix is ​​generated. During implementation, matrix multiplication is used to multiply the weighted matrix with the anomaly vector to calculate the coupling value and expand it into a fluctuation matrix, where each element represents the coupling strength. For example, when the anomaly vector [0.1, 0.2] is multiplied by the weighted matrix to obtain [0.3, 0.29], the dynamic correlation between pressure fluctuations and flow anomalies is revealed. Based on this, by defining a flow anomaly vector and calculating the deviation between the actual flow and its mean, a normalized anomaly vector is obtained. The anomaly vector is then combined with the weighted matrix, and matrix operations are applied to calculate the coupling coefficient, further analyzing the coupling strength between pressure fluctuations and flow anomalies. For example, by combining the matrix row [0.98, 0.97] with the anomaly vector [0.04, -0.04]... The calculation yields a coupling coefficient of 0.0004, revealing a weak coupling relationship and helping to distinguish between normal and abnormal fluctuations. Based on this, an extended matrix is ​​generated using the coupling coefficient. Through threshold filtering, portions with coefficients greater than 0.01 are marked as high coupling, thus forming the final weighted pressure fluctuation matrix. For example, in a zoned pipe network, when the weighted matrix {[0.95, 0.9], [0.9, 0.95]} is calculated with the abnormal vector [0.1, 0.15], the fluctuation matrix [0.2375, 0.2325] is obtained, highlighting high-value areas and helping to improve the sensitivity of leakage detection. This method can not only handle leakage detection in a single area but also be extended to multi-zone scenarios. Through extended matrix calculation, it supports complex pipe network analysis, reduces false alarms, and improves the accuracy of leakage detection.

[0058] S5. Through correlation analysis between the weighted pressure fluctuation matrix and time series features, a feature extraction algorithm is used to determine the candidate location list of leakage points. Specifically, this includes obtaining the pressure fluctuation features in the weighted pressure fluctuation matrix, combining them with the time series features, using the random forest algorithm to extract multi-dimensional feature vectors, calculating the leakage probability of each pipeline node based on the multi-dimensional feature vectors, and sorting them from high to low leakage probabilities to generate a candidate location list of leakage points, which includes the coordinates of multiple candidate nodes.

[0059] Specifically, by acquiring pressure fluctuation characteristics from a weighted pressure fluctuation matrix, this method addresses the problems of inaccurate leakage event identification and difficulty in classifying abnormal patterns in existing technologies. During implementation, the pressure change amplitude and frequency data corresponding to each timestamp are extracted from the weighted pressure fluctuation matrix. Each element of the matrix represents the weighted pressure value of the pipeline within a specific pipeline network section. This matrix is ​​constructed by adjusting the weights of pressure fluctuations through the fusion of pipe diameter ratios. By calculating the extracted fluctuation characteristics, such as average fluctuation amplitude and peak deviation, the physical characteristics of the pipeline network under operating conditions are reflected, providing strong support for abnormal pattern classification. For example, in a city water supply network section, the matrix shows a pressure fluctuation amplitude of 0.5 bar and a frequency of twice per hour, helping to identify potential leaks. This method can be applied to the zoned monitoring of aging pipeline networks. By fusing time-series data of real-time pressure and flow monitoring values, it significantly improves the accuracy of leakage detection and the efficiency of abnormal pattern identification, effectively reducing the false alarm rate in leakage detection.

[0060] By combining time series features and the random forest algorithm to extract multi-dimensional feature vectors, this approach addresses the issues of insufficient accuracy and noise interference in leak detection. In implementation, time series features, such as abnormal flow sequences and pressure fluctuations, are input into the random forest algorithm for processing. The random forest algorithm constructs multiple decision trees and outputs them through voting. Each decision tree is trained from random subsamples to reduce overfitting and improve model robustness. During training, a bootstrapping method is used to sample data and randomly select feature subsets to construct decision trees. The extracted multi-dimensional feature vectors include pressure fluctuation amplitude, flow deviation, and time-series correlation, which helps to more accurately characterize pipeline network sections. By analyzing the abnormal states of points, the accuracy of leakage detection can be improved. For example, in an industrial pipeline network section, the time series shows that the flow rate abnormally exceeds the threshold of 0.3 m³ / h. Combined with pressure fluctuation characteristics, the algorithm extracts a feature vector of [0.4, 0.2, 0.6], representing the abnormal intensity in different dimensions. This processing method helps to reduce noise interference and improve the robustness of features. In another implementation, for pipeline network sections with a high proportion of large-diameter pipes, the algorithm adjusts the tree depth to 15 to extract more time series patterns, adapting to leakage location in complex structures. This method can significantly improve the accuracy of leakage detection, especially in complex pipeline network structures, effectively optimizing the identification of leakage events and pipeline network management.

[0061] This method addresses the challenges of quantifying leakage detection results and inaccurate priority ranking by calculating the leakage probability of each pipeline node based on multi-dimensional feature vectors. In practice, a logistic regression model is used to process each pipeline node, taking the multi-dimensional feature vectors as input and converting a linear combination of the feature vectors into probability values. During training, historical leakage data is used to optimize the model's weights to minimize cross-entropy loss, resulting in a leakage probability score for each node. This approach quantifies the leakage risk of each pipeline node and, combined with prior anomaly pattern classification, further improves leakage detection accuracy. The method improves the accuracy of detection and risk warning capabilities. For example, in a water supply network, when the feature vector of a node is [0.5, 0.3, 0.7], the leakage probability calculated by the logistic regression model is 0.8, indicating that the node has a high probability of leakage. This allows for priority ranking and improved maintenance efficiency. For scenarios with large fluctuations, a regularization term is added to the model to ensure the stability of probability calculation, avoid overfitting, and improve the model's generalization ability. This method can effectively quantify leakage risk, optimize the management and maintenance of water supply networks, and make network leakage detection more accurate and efficient.

[0062] This method generates a candidate list of leakage points based on leakage probabilities, addressing the issues of delayed response and inaccurate location of leakage events. In implementation, the leakage probabilities of all pipeline nodes are first sorted, and the top 10 high-probability nodes are selected from highest to lowest probability. Then, the geographical coordinates of these nodes are correlated with their probabilities to form a candidate list of leakage points. This list contains the coordinates of multiple candidate nodes, providing a basis for precise location of leakage events and ensuring rapid response. For example, in a water supply network scenario, the sorted list includes node A (probability 0.9, coordinates (120.5, 30.2)) and node B (probability 0.85, coordinates (120.6, 30.3)) as candidate locations, helping pipeline operators to locate and address leakage problems promptly. This method effectively improves the response speed and location accuracy of leakage events.

[0063] S6. Obtain neighborhood connection data from the candidate location list of leak points, and determine high-probability leak points based on the matching degree to obtain a precise location coordinate set. Specifically, this includes obtaining the neighborhood connection data of each candidate node in the candidate location list of leak points, calculating the matching degree between the neighborhood connection data and the multi-dimensional feature vector, and determining the corresponding candidate node as a high-probability leak point if the matching degree is greater than or equal to a preset matching threshold, and determining the corresponding candidate node as a low-probability leak point if the matching degree is less than the preset matching threshold. Based on the coordinates of the high-probability leak points, generate a precise location coordinate set, which includes the spatial location of the high-probability leak points.

[0064] Specifically, by acquiring neighborhood connection data from the candidate location list of leak points and combining it with matching degree calculation, leak points are located, solving problems such as inaccurate leak location and slow response. During implementation, the pipeline topology database is queried to obtain the connecting pipe information around each candidate node, including the neighboring node number, pipe length, and connection type, forming a neighborhood connection data structure. Then, the matching degree between the neighborhood connection data and the multi-dimensional feature vector extracted from the random forest algorithm is calculated. Specifically, the neighborhood connection data is converted into vector form, and the connection relationship is represented using a graph embedding method. Then, the similarity between the neighborhood connection data vector and the multi-dimensional feature vector is calculated using cosine similarity to obtain the matching degree value. Cosine similarity measures the angular difference between two vectors by dividing their dot product by their magnitude. For example, if the multi-dimensional feature vector of a candidate node is [0.8, 0.5, 0.3], corresponding to pressure fluctuation, flow anomaly, and pipe diameter ratio, and the embedding vector of the neighborhood connection data is [0.7, 0.6, 0.4], the calculated cosine similarity is 0.95, indicating a high match. In one implementation, for industrial pipeline networks such as oil pipelines, the matching degree calculation further incorporates time series factors. This is achieved by weighting the neighboring connection data over time and calculating the Euclidean distance to the feature vector to supplement the matching degree calculation. The Euclidean distance quantifies the straight-line distance between vectors, making the model more robust in dynamic flow change scenarios. For example, if the feature vector of a node in an industrial pipeline network is [1.2, 0.9, 0.6] and the connection data vector is [1.1, 0.8, 0.7], the calculated Euclidean distance is 0.17. A low distance indicates a high matching degree, which is beneficial for quickly responding to potential leaks and reducing economic losses. If the calculated matching degree is higher than a preset matching threshold, such as 0.8, the node is marked as a high-probability leak point. Finally, based on the coordinates of these high-probability leak points, a precise location coordinate set is generated to support leak event path generation and precise location. This method effectively improves the accuracy of leak detection, optimizes pipeline network management and maintenance efficiency, and is particularly suitable for leak location and response in complex pipeline structures.

[0065] S7. Generate a dynamic correlation graph based on the precise positioning coordinate set to obtain the propagation path of leakage events; S8. Optimize the path weights of the dynamic correlation graph using an optimization algorithm to determine the final leakage point; Specifically, this includes constructing the topology of the pipeline network based on the precise positioning coordinate set, generating a dynamic correlation graph of propagation paths through the topology, integrating the event propagation paths in the dynamic correlation graph, calculating the propagation weights of each path, and determining the propagation path of leakage events based on the propagation weights, where the propagation path of leakage events represents the range of influence of the leakage point on the operating status of the pipeline network.

[0066] Specifically, by constructing a pipeline network topology, this method addresses the issues of insufficient integration of leakage point information and inefficient path analysis. During implementation, coordinates from a precisely located coordinate set are mapped onto pipeline nodes and connections, forming node sets and edge sets. Node sets represent pipeline intersections, while edge sets represent pipeline segments. A graph model of the pipeline network is established by combining these sets. This graph model clearly represents the connections between each pipeline intersection and adjacent pipeline segments, providing a network foundation for path analysis of leakage events. More specifically, the precise location coordinate set allows for rapid integration of leakage point location information, forming a pipeline network topology that facilitates path generation and leakage location. This method improves the efficiency of leakage detection, enabling more accurate and rapid pipeline network management and maintenance, particularly in complex pipeline systems, where it facilitates efficient leakage event analysis and rapid response.

[0067] This method addresses the issues of inaccurate propagation analysis and slow response in leak detection by generating a dynamic correlation graph of propagation paths. During implementation, potential propagation paths are extracted from the pipeline network topology and integrated with real-time monitoring data such as pressure and flow rates to construct dynamic correlations between paths. For each path, the correlation strength is calculated, and a graph-based adjacency matrix is ​​used to represent the dynamic changes of the paths. By updating the correlation graph, time-series data is incorporated to reflect the real-time propagation dynamics of leak events. The dynamic correlation graph traverses paths using a depth-first search algorithm, marking highly correlated areas, and its structure is optimized to ensure it can capture the complex connections within the pipeline network. This method effectively identifies the dynamic propagation characteristics of leak events in pipeline networks, especially in older networks such as urban water supply networks. In this scenario, by integrating pipe diameter ratio and pressure fluctuation data, the map can be updated with path associations in real time, thereby improving the accuracy of leakage detection. For example, in a zoned pipe network, if the coordinate set accurately points to a potential leakage point, when generating the map through the topology, considering neighborhood connection data, the path association strength can reach above 0.8, which can identify the propagation path affecting downstream areas, reduce false positives, and improve response efficiency. In complex pipe network structures, especially in scenarios with pipes of different diameters, the map can be extended to multi-level generation. By adjusting the weight of pipe diameters, the association relationships in the map can be dynamically adjusted to ensure that the propagation path from the main road to the branch pipes is fully covered. Through this method, leakage events in the pipe network can be captured more comprehensively and accurately, optimizing the dynamic management and maintenance of the water supply network and improving the accuracy and efficiency of leakage detection.

[0068] By integrating event propagation paths from a dynamic correlation graph and calculating the propagation weight of each path, the problems of inaccurate analysis and unclear priority determination of leakage event propagation paths can be effectively solved. In implementation, the event propagation path sequence is first extracted from the dynamic correlation graph and then integrated with the time series features of the preliminary leakage signal set to calculate the propagation weight of each path. Specifically, a weighted average method is used to integrate the pressure fluctuation matrix and flow anomalies to calculate the weight factor of each path. The weight factor is defined as the ratio of path length to anomaly intensity; for example, if the path length is L and the anomaly intensity is I, then the weight factor is L / I. Subsequently, a web-level algorithm is applied to simulate the propagation weight. The importance of the path is calculated iteratively. The web-level algorithm uses the in-degree and out-degree of nodes to evaluate the importance of the path and updates the propagation weight of the path based on the weight transfer of adjacent paths. This method yields the propagation weight value of each path, serving as a quantitative indicator of the event's impact. Furthermore, combined with pipe... The weights of network physical characteristics, such as pipe diameter ratio, are adjusted to further improve the accuracy of weight calculation. For example, in aging pipe network zones, the propagation weight calculated by integrating real-time pressure data and pipe diameter weights can indicate high-risk paths, helping to identify potential expansion areas that need to be prioritized and reduce overall losses. Assuming an initial weight of 0.2 for a path, its propagation weight increases to 0.35 after iteration using a web-level algorithm. Combined with an abnormal flow exceeding the threshold of 0.5 cubic meters per hour, this path is confirmed as a critical propagation chain, which helps optimize maintenance strategies. For scenarios with complex pipe network structures, the weight calculation can be extended to multi-dimensional weight calculation. By incorporating neighborhood connection matching degree, if the matching degree is higher than 0.7, the weight is increased by 20%, ensuring the robustness of weight calculation and adapting to different situations from single leaks to chain reactions. This method can effectively improve the accuracy of propagation path analysis of leak events, optimize the response strategy for leak detection, and improve the management and maintenance efficiency of water supply networks.

[0069] This method addresses the issues of unclear impact range identification and difficulty in accurately planning pipeline maintenance by determining the propagation paths of leakage events. During implementation, paths with weights exceeding a preset threshold are selected to form a set of leakage event propagation paths. By evaluating this set, the impact range of leakage events on the pipeline network's operational status is further determined, including affected nodes and areas, thus achieving precise analysis of leakage event propagation. This method clearly characterizes the impact of leakage points on the pipeline network, helps identify potential risk areas, and provides targeted planning and optimization for pipeline maintenance. Furthermore, this method effectively improves leakage detection and response strategies, ensuring rapid and accurate handling of leakage events in the pipeline network and optimizing the operation and management of the water supply network.

[0070] The above describes a method for dynamic analysis of leakage in a water supply network using DMA partitioning, as described in an embodiment of this application. The following describes a system for dynamic analysis of leakage in a water supply network using DMA partitioning, as described in an embodiment of this application. Please refer to [link to relevant documentation]. Figure 2 One embodiment of the DMA zone leakage dynamic analysis system for water supply networks in this application includes:

[0071] The data collection unit is used to acquire pipeline data and real-time monitoring data within the DMA partition of the water supply network, and to obtain the partition's physical characteristic indicators and operating status vector.

[0072] The event determination unit is used to determine the sequence of potential leakage events based on partition physical characteristic indicators and operating status vectors using a classification algorithm.

[0073] The feature extraction unit is used to extract time series features from the potential leakage event sequence, and combine them with a preset threshold to judge the preliminary leakage signal and generate a preliminary leakage signal set.

[0074] The matrix construction unit is used to construct a weighted matrix by fusing pipeline data with the initial set of leakage signals, thereby generating a weighted pressure fluctuation matrix.

[0075] The correlation analysis unit is used to determine the list of candidate locations of leakage points by using feature extraction algorithms through correlation analysis between the weighted pressure fluctuation matrix and time series features.

[0076] The location matching unit is used to obtain neighborhood connection data from the candidate location list of leakage points, and combine the matching degree to determine the high probability of leakage points, so as to obtain the precise location coordinate set.

[0077] The map generation unit is used to generate a dynamic correlation map based on the precise positioning coordinate set, thereby obtaining the propagation path of leakage events.

[0078] The path optimization unit is used to optimize the path weights of the dynamic association graph through optimization algorithms to determine the final leakage points.

[0079] Through the collaborative efforts of the aforementioned components, the detection accuracy and location efficiency of leakage events have been further improved.

[0080] This application also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of a dynamic analysis method for leakage in a water supply network DMA partition.

[0081] In summary, this application addresses the problems of low leakage detection accuracy and difficulty in tracing the propagation path of leakage events in existing technologies by integrating multi-dimensional data and dynamic analysis. It employs a dynamic correlation map based on the pipeline network topology and machine learning algorithms to extract precise leakage signals from real-time monitoring data of the water supply network. Combined with the physical characteristics of the pipeline network and dynamic data analysis, it accurately locates leakage points and tracks their propagation paths in real time. By combining a weighted matrix with a depth-first search algorithm, this application can not only accurately identify potential leakage areas when leakage occurs, but also dynamically adjust path weights under complex pipeline network structures, optimizing leakage event repair and pipeline maintenance strategies.

[0082] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0083] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0084] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A dynamic analysis method for leakage in a water supply network by DMA (Districtive Displacement) zone, characterized in that, The method includes: S1. Obtain pipeline data and real-time monitoring data within the DMA partition of the water supply network to obtain the partition's physical characteristic indicators and operating status vector; S1 further includes: collecting diameter distribution ratio data, real-time pressure data, and flow monitoring data of pipes with different diameters within the water supply network DMA partition; fusing time-series pressure data with pipe diameter ratio data; calculating pipe diameter weights; adjusting the real-time pressure data and flow monitoring data based on the pipe diameter weights; and generating partition physical characteristic indicators; combining the partition physical characteristic indicators with the real-time monitoring data to construct an operating state vector, wherein the operating state vector characterizes the dynamic operating characteristics of the water supply network DMA partition. S2. Based on the partition physical characteristic indicators and the running state vector, a classification algorithm is used to determine the sequence of potential leakage events; The S2 further includes: inputting the partition physical characteristic index and the running state vector into a support vector machine algorithm, performing abnormal pattern classification through the support vector machine algorithm to obtain a classification result, calculating the fluctuation amplitude of the running state vector based on the classification result, and if the fluctuation amplitude exceeds a preset fluctuation threshold, determining the corresponding running state vector as a potential leakage event and generating a potential leakage event sequence. S3. Extract time series features from the potential leakage event sequence, and combine them with a preset threshold to determine the preliminary leakage signal, and generate a preliminary leakage signal set; S3 further includes: extracting time series features from the potential leakage event sequence, wherein the time series features include flow rate change rate and pressure fluctuation frequency; based on the time series features, determining whether the flow anomaly exceeds a preset flow threshold; if the flow anomaly exceeds the preset flow threshold, then combining the anomaly pattern classification results to determine the preliminary leakage signal; and summarizing all the preliminary leakage signals to generate a preliminary leakage signal set. S4. For the initial set of leakage signals, fuse pipeline data to construct a weighted matrix and generate a weighted pressure fluctuation matrix; S4 further includes: acquiring pressure fluctuation data from the preliminary leakage signal set, fusing pipe diameter ratio data, adjusting the weights of the pressure fluctuation data to generate a weighted pressure value, constructing a weighted matrix based on the correlation between the weighted pressure value and the flow monitoring data, calculating the coupling relationship between pressure fluctuation and flow anomaly through the weighted matrix, and generating a weighted pressure fluctuation matrix. S5. Through the correlation analysis between the weighted pressure fluctuation matrix and the time series features, a feature extraction algorithm is used to determine the list of candidate locations for leakage points; S6. Obtain neighborhood connection data from the candidate location list of the leakage point, and determine the high probability leakage point by combining the matching degree to obtain the precise location coordinate set; S7. Generate a dynamic correlation map based on the precise positioning coordinate set to obtain the propagation path of the leakage event; S8. Optimize the path weights of the dynamic association graph through optimization algorithms to determine the final leakage points.

2. The method according to claim 1, characterized in that, The S5 also includes: The pressure fluctuation features in the weighted pressure fluctuation matrix are obtained, and combined with the time series features, a multi-dimensional feature vector is extracted using the random forest algorithm. Based on the multi-dimensional feature vector, the leakage probability of each pipeline node is calculated, and the leakage probabilities are sorted from high to low to generate a list of candidate leakage points, wherein the list of candidate leakage points includes the coordinates of multiple candidate nodes.

3. The method according to claim 1, characterized in that, S6 further includes: Obtain the neighborhood connection data of each candidate node in the candidate location list of the leakage point, calculate the matching degree between the neighborhood connection data and the multi-dimensional feature vector, if the matching degree is greater than or equal to a preset matching threshold, then the corresponding candidate node is determined as a high-probability leakage point, if the matching degree is less than the preset matching threshold, then the corresponding candidate node is determined as a low-probability leakage point, and based on the coordinates of the high-probability leakage point, generate a precise positioning coordinate set, wherein the precise positioning coordinate set includes the spatial location of the high-probability leakage point.

4. The method according to claim 1, characterized in that, The S7 also includes: Based on the precise positioning coordinate set, the topology of the pipeline network is constructed, and a dynamic correlation graph of propagation paths is generated through the topology. Event propagation paths in the dynamic correlation graph are merged, and the propagation weight of each path is calculated. Based on the propagation weight, the propagation path of leakage events is determined, wherein the propagation path of leakage events represents the range of influence of leakage points on the operating status of the pipeline network.

5. A dynamic analysis system for DMA-based leakage in a water supply network, used to implement the dynamic analysis method for DMA-based leakage in a water supply network as described in any one of claims 1-4, characterized in that, The system includes: The data collection unit is used to acquire pipeline data and real-time monitoring data within the DMA partition of the water supply network, and to obtain the partition's physical characteristic indicators and operating status vector. The event determination unit is used to determine a sequence of potential leakage events based on the partition physical characteristic indicators and the running state vector using a classification algorithm; The feature extraction unit is used to extract time series features from the potential leakage event sequence, and combine them with a preset threshold to determine the preliminary leakage signal and generate a preliminary leakage signal set. The matrix construction unit is used to construct a weighted matrix by fusing pipeline data with the preliminary leakage signal set, and generate a weighted pressure fluctuation matrix. The correlation analysis unit is used to determine a list of candidate locations for leakage points by using a feature extraction algorithm through correlation analysis between the weighted pressure fluctuation matrix and the time series features. The location matching unit is used to obtain neighborhood connection data from the list of candidate locations of the leak points, and combine the matching degree to determine the high probability of leak points to obtain a precise location coordinate set; The map generation unit is used to generate a dynamic correlation map based on the precise positioning coordinate set to obtain the propagation path of the leakage event. The path optimization unit is used to optimize the path weights of the dynamic association graph through optimization algorithms to determine the final leakage points.

6. A computer-readable storage medium storing instructions thereon, characterized in that, When the instruction is executed by the processor, it implements a dynamic analysis method for leakage in a water supply network DMA partition as described in any one of claims 1-4.

Images