A water supply network leakage positioning method and system

Abstract

This invention relates to the field of data processing technology and discloses a method and system for locating leaks in water supply networks. The method includes: performing adaptive filtering and robust outlier cleaning on real-time pressure monitoring data of the water supply network to obtain a cleaned pressure data sequence; fusing the multi-scale permutation entropy and instantaneous energy spectrum of the cleaned pressure data sequence into a high-dimensional signal fingerprint feature vector; mapping the topological structure data into a directed weighted graph of the water supply network and injecting the high-dimensional signal fingerprint feature vector into the directed weighted graph to obtain a causal attribute graph; performing pressure anomaly deduction on the causal attribute graph to obtain potential propagation paths and a set of root source nodes; using the nodes in the root source node set as indexes, traversing backwards through the associated branches of the potential propagation paths to accurately locate the leakage interval of the water supply network; and comprehensively evaluating the leakage interval to obtain a leakage location diagnosis report. This invention can improve the efficiency of locating leaks in water supply networks.

Description

Technical Field

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

[0002] Existing technologies have significant shortcomings in the data preprocessing and feature extraction stages for locating leaks in water supply networks. They fail to perform adaptive filtering and robust outlier removal on real-time pressure monitoring data, relying instead on simple filtering or directly using raw data. This results in severe noise interference and ineffective outlier removal, making it impossible to form accurate pressure data sequences. Furthermore, they fail to fuse multi-scale permutation entropy with instantaneous energy spectrum to construct a high-dimensional signal fingerprint feature vector, relying solely on single-dimensional features or simple combinations of features. This makes it difficult to comprehensively capture the complex changing patterns of pressure signals, leading to insufficient representation of leak-related features and creating potential errors for subsequent location work.

[0003] Existing technologies have significant shortcomings in the mapping and location inference stages of water supply network leakage localization. They fail to map topological data into a directed weighted graph and inject feature vectors to form a causal attribute graph; instead, they rely solely on isolated topological or feature data, failing to establish a correlation between network structure and pressure anomalies. Furthermore, they fail to use the causal attribute graph to infer potential propagation paths and root cause node sets for pressure anomalies, relying instead on blindly traversing the network or depending on experience, making it difficult to pinpoint core nodes and propagation links related to leakage. Finally, they fail to combine similarity clustering results with reverse traversal of associated branches to accurately locate leakage intervals, only roughly dividing suspected areas, resulting in low positioning accuracy and ambiguous ranges, making it impossible to quickly and efficiently locate leaks and failing to meet the actual need for timely repair of water supply network leaks. Summary of the Invention

[0004] This invention provides a method and system for locating leaks in water supply networks to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a method for locating leakage in a water supply network, comprising:

[0006] S1. Adaptive filtering and robust outlier cleaning are performed on the real-time pressure monitoring data of the water supply network to obtain the cleaned pressure data sequence of the water supply network;

[0007] S2. The multi-scale permutation entropy of the pressure data sequence after cleaning and the instantaneous energy spectrum are fused into a high-dimensional signal fingerprint feature vector of the water supply network;

[0008] S3. Map the topological data of the water supply network to a directed weighted graph of the water supply network, and inject the high-dimensional signal fingerprint feature vector into the directed weighted graph to obtain the causal attribute graph of the water supply network;

[0009] S4. Perform pressure anomaly deduction on the causal attribute map to obtain the potential propagation path and root cause node set of the water supply network;

[0010] S5. Using the nodes in the root node set as indexes, traverse the associated branches of the potential propagation path in reverse, and combine the similarity clustering results of the associated branches to accurately locate the leakage range of the water supply network.

[0011] S6. A comprehensive assessment of the leakage area is conducted to obtain a leakage location diagnosis report for the water supply network.

[0012] In a preferred embodiment, the adaptive filtering and outlier robust cleaning of the real-time pressure monitoring data of the water supply network to obtain the cleaned pressure data sequence of the water supply network includes:

[0013] Receive real-time pressure monitoring data of the water supply network to obtain the original pressure sequence of the water supply network;

[0014] The original pressure sequence is adaptively filtered to obtain the filtered pressure sequence of the water supply network.

[0015] Outlier identification is performed on the filtered pressure sequence to obtain the anomaly marker sequence of the water supply network;

[0016] Based on the anomaly marker sequence, the filtered pressure sequence is cleaned to obtain the cleaned pressure data sequence of the water supply network.

[0017] In a preferred embodiment, the step of fusing the multi-scale permutation entropy of the post-cleaning pressure data sequence with the instantaneous energy spectrum to form a high-dimensional signal fingerprint feature vector for the water supply network includes:

[0018] Multi-scale coarse-grained analysis was performed on the pressure data sequence after cleaning to obtain the coarse-grained sequence of the water supply network;

[0019] The coarse-grained sequence is topologically reconstructed to obtain the time-delay phase space of the water supply network;

[0020] In the time-delay phase space, the arrangement pattern of the coarse-grained sequence is identified to statistically determine the distribution probability of the coarse-grained sequence;

[0021] The information entropy quantization process is performed on the probability distribution to obtain the multi-scale arrangement entropy of the water supply network.

[0022] In a preferred embodiment, the step of fusing the multi-scale permutation entropy of the post-cleaning pressure data sequence with the instantaneous energy spectrum to form a high-dimensional signal fingerprint feature vector for the water supply network includes:

[0023] The time-frequency representation of the pressure data sequence after cleaning is obtained by performing a time-frequency transformation on the pressure data sequence after cleaning.

[0024] The instantaneous energy spectrum of the water supply network is obtained by characterizing the energy value of the time-frequency representation of the signal.

[0025] The instantaneous energy spectrum and the multi-scale arrangement entropy are vectorized and concatenated to obtain the preliminary fingerprint feature vector of the water supply network;

[0026] The preliminary fingerprint feature vector is normalized and integrated to obtain the high-dimensional signal fingerprint feature vector of the water supply network.

[0027] In a preferred embodiment, reconstructing the topology data of the water supply network into a directed weighted graph of the water supply network includes:

[0028] Collect the topology data of the water supply network;

[0029] Using the topology data as a reference, the connection relationships of the network components representing the water supply network are determined;

[0030] Based on the aforementioned characterization of the network component connection relationship, the network components of the topology data are defined as node groups;

[0031] Based on the connection relationship of the network components, a directed connection is made between the node groups to obtain the directed configuration connection group of the water supply network.

[0032] Based on the pipe segment attributes in the topology data, the directed connection groups are assigned corresponding connection weights.

[0033] Based on the connection weights, the node groups are sequentially connected according to the directed connection groups to construct a directed weighted graph of the water supply network.

[0034] In a preferred embodiment, injecting the high-dimensional signal fingerprint feature vector into the directed weighted graph to obtain the causal attribute graph of the water supply network includes:

[0035] The node location information in the topology data is spatially aligned with the acquisition location information of the real-time pressure monitoring data to obtain the spatial mapping relationship of the water supply network.

[0036] The time synchronization relationship of the water supply network is obtained by calibrating the timestamp of the real-time pressure monitoring data with the generation time of the high-dimensional signal fingerprint feature vector.

[0037] Based on the spatial mapping relationship and the time synchronization relationship, a multi-dimensional correlation mapping analysis is performed on the high-dimensional signal fingerprint feature vector and the corresponding monitoring node in the directed weighted graph to obtain the accurate matching relationship of the water supply network.

[0038] Based on the precise matching relationship, the high-dimensional signal fingerprint feature vector is used as an attribute feature and assigned to the corresponding node in the directed weighted graph to obtain the causal attribute graph of the water supply network.

[0039] In a preferred embodiment, the step of performing pressure anomaly deduction on the causal attribute map to obtain the potential propagation paths and root cause node set of the water supply network includes:

[0040] Extract the node attributes of the high-dimensional signal fingerprint feature vector from the nodes of the causal attribute graph;

[0041] Based on the node attributes, assess the degree of pressure state anomaly of the nodes in the causal attribute graph;

[0042] Based on the degree of pressure anomaly, the directed propagation process between the nodes is simulated to obtain the root node set of the water supply network.

[0043] By tracing the state propagation beginning of the root node, the resulting state propagation link is determined as the potential propagation path of the water supply network.

[0044] In a preferred embodiment, the step of using nodes in the root node set as indexes to traverse the associated branches of the potential propagation path in reverse, and combining the similarity clustering results of the associated branches to accurately locate the leakage range of the water supply network includes:

[0045] Based on the degree of pressure anomaly of the root node on the potential propagation path, the reverse traversal priority for the potential propagation path is dynamically allocated;

[0046] Based on the reverse traversal priority, a progressive traversal is performed in the opposite direction of the connection in the directed weighted graph, and the topological connection structure and signal fingerprint features of the nodes passed through are aggregated in real time during the traversal process to obtain the candidate associated branches of the water supply network.

[0047] Based on the candidate association branches, construct the fusion similarity index of the water supply network;

[0048] Based on the fusion similarity index, all candidate association branches are grouped and clustered to obtain the clustering groups of the water supply network;

[0049] From the clustering groups, the group with the highest fusion similarity index and the most significant abnormal performance of the signal fingerprint features is identified as the core abnormal group of the water supply network;

[0050] Based on the core anomaly group, the continuous pipe sections of the water supply network are determined to obtain the leakage range of the water supply network.

[0051] In a preferred embodiment, the formula for calculating the fusion similarity index is as follows:

[0052] ;

[0053] In the formula, The fusion similarity index, The topological similarity components between different branches in the candidate association branches. The feature similarity components between different branches in the candidate association branches. This refers to the dynamic adjustment coefficient of the candidate association branch. Let A be the topological depth of branch A among the candidate related branches. Let B be the topological depth of the candidate associated branches. The branch normalization factor is the number of candidate related branches. It is an exponential function.

[0054] To address the above problems, the present invention also provides a water supply network leakage location system, the system comprising:

[0055] A standard cleaning module is used to perform adaptive filtering and robust outlier cleaning on real-time pressure monitoring data of the water supply network to obtain a pressure data sequence of the water supply network after cleaning.

[0056] The fusion vector module is used to fuse the multi-scale permutation entropy of the pressure data sequence after cleaning with the instantaneous energy spectrum into a high-dimensional signal fingerprint feature vector of the water supply network.

[0057] An image construction module is used to reconstruct the topological data of the water supply network into a directed weighted graph of the water supply network, and inject the high-dimensional signal fingerprint feature vector into the directed weighted graph to obtain the causal attribute graph of the water supply network.

[0058] Anomaly simulation module is used to perform pressure anomaly simulation on the causal attribute map to obtain the potential propagation path and root cause node set of the water supply network.

[0059] The leakage location module is used to traverse the associated branches of the potential propagation path in reverse using the nodes in the root node set as indexes, and combine the similarity clustering results of the associated branches to accurately locate the leakage range of the water supply network.

[0060] The location diagnosis module is used to comprehensively analyze the leakage range and obtain a leakage location diagnosis report for the water supply network.

[0061] Compared with the prior art, the present invention has the following beneficial effects:

[0062] 1. This invention provides high-quality data support for locating leaks in water supply networks through precise data processing and high-dimensional feature fusion. Adaptive filtering and outlier cleaning are applied to real-time pressure monitoring data to obtain a clean pressure data sequence. A high-dimensional signal fingerprint feature vector is constructed by fusing multi-scale permutation entropy and instantaneous energy spectrum, comprehensively capturing the complex changing characteristics of the pressure signal and providing accurate evidence for leak identification.

[0063] 2. This invention significantly improves the efficiency and accuracy of leak location by leveraging graph construction and precise inference and localization. Topological data is mapped into a directed weighted graph and feature vectors are injected to form a causal attribute graph, from which potential propagation paths and root node sets are derived. Using root node as an index, related branches are traversed in reverse, and similarity clustering is combined to accurately locate leak intervals. A diagnostic report is generated through comprehensive analysis, achieving rapid and accurate leak location. Attached Figure Description

[0064] Figure 1 This is a flowchart illustrating a method for locating leakage in a water supply network according to an embodiment of the present invention.

[0065] Figure 2 A functional block diagram of a water supply network leakage location system provided in an embodiment of the present invention;

[0066] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0067] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0068] This application provides a method for locating leaks in a water supply network. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for locating leaks in a water supply network can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0069] Reference Figure 1 The diagram shown is a flowchart illustrating a method for locating leaks in a water supply network according to an embodiment of the present invention. In this embodiment, the method for locating leaks in a water supply network includes:

[0070] S1. Adaptive filtering and robust outlier cleaning are performed on the real-time pressure monitoring data of the water supply network to obtain the cleaned pressure data sequence of the water supply network;

[0071] In this embodiment of the invention, the adaptive filtering and robust outlier cleaning of the real-time pressure monitoring data of the water supply network to obtain the cleaned pressure data sequence of the water supply network includes:

[0072] Receive real-time pressure monitoring data of the water supply network to obtain the original pressure sequence of the water supply network;

[0073] The original pressure sequence is adaptively filtered to obtain the filtered pressure sequence of the water supply network.

[0074] Outlier identification is performed on the filtered pressure sequence to obtain the anomaly marker sequence of the water supply network;

[0075] Based on the anomaly marker sequence, the filtered pressure sequence is cleaned to obtain the cleaned pressure data sequence of the water supply network.

[0076] Pressure sensors deployed in the water supply network continuously collect pressure data during the network's operation. These data are arranged sequentially according to the time of collection, forming a continuous data sequence that can completely reflect the real-time pressure changes of the network, thus obtaining the original pressure sequence of the water supply network.

[0077] Based on the variation amplitude and fluctuation frequency of pressure data in the original pressure sequence, the size and intensity of the filtering window are dynamically adjusted. When the data fluctuates drastically, the window is enlarged and the filtering intensity is increased; when the data is relatively stable, the window is reduced and the filtering intensity is weakened. According to the adjusted filtering parameters, each data point in the original pressure sequence is smoothed to eliminate high-frequency interference and irregular fluctuations in the data, while retaining the core trend of pressure changes, thus obtaining the filtered pressure sequence of the water supply network.

[0078] The pressure fluctuation range under normal operating conditions of the water supply network is set as the anomaly judgment standard. This standard is determined based on the network's design pressure parameters, historical operating data, and safety operation requirements. Each data point in the filtered pressure sequence is checked one by one to determine whether it is within the preset normal pressure fluctuation range. If a data point exceeds this range, it is marked as abnormal data. The judgment results of all data points are arranged in the corresponding order to obtain the anomaly marking sequence of the water supply network.

[0079] Based on the marking results in the anomaly marker sequence, the locations of abnormal data in the filtered pressure sequence are identified. The abnormal data is then replaced using the mean of adjacent normal data to ensure that the replaced data maintains the same trend as the surrounding normal data. The pressure sequence after the replacement process is then validated to confirm that no abnormal data is missing, the data logic is coherent, and the fluctuations are reasonable. This results in the final pressure data sequence after the water supply network cleaning.

[0080] The beneficial effects are that it receives real-time pressure monitoring data from the water supply network and forms a raw pressure sequence, fully capturing the pressure changes during the operation of the network, providing comprehensive and tamper-proof basic data for subsequent data processing, and ensuring that leakage location work has a reliable data source.

[0081] Adaptive filtering is applied to the original pressure sequence, and the filtering parameters are dynamically adjusted according to the data fluctuation characteristics to accurately eliminate high-frequency interference and irregular fluctuations, while retaining the core trend of pressure changes. This results in a filtered pressure sequence, which improves the purity and stability of the data and provides accurate data support for subsequent anomaly identification.

[0082] Based on a preset normal pressure fluctuation range, outlier identification is performed on the filtered pressure sequence. Abnormal data that exceeds the reasonable range is accurately marked, forming an anomaly marker sequence. This clarifies the target object of data cleaning and avoids interference from abnormal data in the extraction of loss features.

[0083] The filtered pressure sequence is cleaned based on the anomaly marker sequence. By using a reasonable method to replace the anomaly data, the replacement data is made to ensure that the trend of the replaced data is consistent with that of the surrounding normal data. Finally, the cleaned pressure data sequence is obtained, which ensures the continuity, accuracy and usability of the data and lays a solid foundation for the construction of high-dimensional signal fingerprint feature vectors.

[0084] S2. The multi-scale permutation entropy of the pressure data sequence after cleaning and the instantaneous energy spectrum are fused into a high-dimensional signal fingerprint feature vector of the water supply network;

[0085] In this embodiment of the invention, the step of fusing the multi-scale permutation entropy of the post-cleaning pressure data sequence with the instantaneous energy spectrum to form a high-dimensional signal fingerprint feature vector for the water supply network includes:

[0086] Multi-scale coarse-grained analysis was performed on the pressure data sequence after cleaning to obtain the coarse-grained sequence of the water supply network;

[0087] The coarse-grained sequence is topologically reconstructed to obtain the time-delay phase space of the water supply network;

[0088] In the time-delay phase space, the arrangement pattern of the coarse-grained sequence is identified to statistically determine the distribution probability of the coarse-grained sequence;

[0089] The information entropy quantization process is performed on the probability distribution to obtain the multi-scale arrangement entropy of the water supply network.

[0090] The process of fusing the multi-scale permutation entropy of the post-cleaning pressure data sequence with the instantaneous energy spectrum to form a high-dimensional signal fingerprint feature vector for the water supply network includes:

[0091] The time-frequency representation of the pressure data sequence after cleaning is obtained by performing a time-frequency transformation on the pressure data sequence after cleaning.

[0092] The instantaneous energy spectrum of the water supply network is obtained by characterizing the energy value of the time-frequency representation of the signal.

[0093] The instantaneous energy spectrum and the multi-scale arrangement entropy are vectorized and concatenated to obtain the preliminary fingerprint feature vector of the water supply network;

[0094] The preliminary fingerprint feature vector is normalized and integrated to obtain the high-dimensional signal fingerprint feature vector of the water supply network.

[0095] Following a fixed scaling rule, the post-cleaning pressure data sequence is divided into several consecutive data segments of equal length in chronological order, with each segment serving as an independent coarse-grained processing unit. The mean value of all pressure data within each segment is calculated, and this mean is used as the representative value of that coarse-grained unit. These representative values ​​are then arranged sequentially according to the original data segments' chronological order, forming a simplified data sequence that reflects pressure variation characteristics at different scales, thus obtaining the coarse-grained sequence of the water supply network.

[0096] A fixed time delay parameter is determined, based on the variation cycle and historical patterns of the water supply network pressure data, to ensure accurate capture of the temporal correlation between data points. Using each data point in the coarse-grained sequence as an initial reference point, subsequent corresponding related data points are extracted sequentially according to the set time delay parameter. Each initial reference point and its corresponding related data point are combined to form a vector in a high-dimensional space. All generated vectors are arranged in the order of the initial reference points to construct a spatial structure that reflects the temporal evolution and inherent correlation of the coarse-grained sequence, thus obtaining the time delay phase space of the water supply network.

[0097] Iterate through each vector in the time-delay phase space, sort all data points within each vector by value, and record the positional order of the data points in the original vector to form the permutation pattern corresponding to that vector. Compare the permutations of all vectors one by one, count the frequency of each permutation pattern in the entire time-delay phase space, and compare the frequency of each permutation pattern with the total number of vectors to obtain the proportion of each permutation pattern in the whole. This proportion is the probability distribution of the coarse-grained sequence.

[0098] Collect the probability distributions corresponding to all permutation patterns, perform a logarithmic transformation on each probability distribution, and then multiply the transformed result by the corresponding probability distribution to obtain the information contribution value of each permutation pattern. Sum the information contribution values ​​of all permutation patterns, and then take the negative of the summation result. In this way, the probability distribution is transformed into an entropy value that can quantify the complexity of the sequence. This entropy value is the multi-scale permutation entropy of the water supply network.

[0099] A fixed time-frequency transformation method is used to transform the pressure data sequence after cleaning from the time domain to the time-frequency joint domain. By decomposing the time variation characteristics and frequency distribution characteristics of the data sequence, a two-dimensional representation form that can simultaneously reflect the intensity distribution of pressure data at different time points and different frequency bands is obtained. This representation form is the time-frequency representation of the pressure data sequence after cleaning.

[0100] For each time-frequency coordinate point in the signal time-frequency representation, the squared value of the signal amplitude corresponding to that coordinate point is calculated. This value is used as the energy quantization result for that point, comprehensively covering all coordinate points in the signal time-frequency representation to ensure that no energy information under any time and frequency combination is missed. The energy quantization results of all coordinate points are then arranged in chronological and frequency order to form a continuous spectrum that dynamically reflects the energy changes of the water supply network pressure signal at different times and frequencies, thus obtaining the instantaneous energy spectrum of the water supply network.

[0101] All energy data in the instantaneous energy spectrum are expanded into a one-dimensional energy vector in a preset order. At the same time, the quantized value of the multi-scale permutation entropy is retained as an independent feature component. The two are connected end to end in a fixed order of first energy vector and then permutation entropy component, so that energy features and complexity features are organically combined to form a combined vector that contains both time-frequency energy information and sequence complexity information, thus obtaining the preliminary fingerprint feature vector of the water supply network.

[0102] The maximum and minimum values ​​in the preliminary fingerprint feature vector are determined, and the difference between them is calculated as the fluctuation range of the vector data. The minimum value of the vector is subtracted from each feature component in the preliminary fingerprint feature vector, and then divided by the fluctuation range. Through this unified scaling process, all feature components are mapped to a fixed numerical range, eliminating the influence of dimensional differences and numerical spans between different types of features, so that each feature component in the vector has a unified comparability basis, and finally the high-dimensional signal fingerprint feature vector of the water supply network is obtained.

[0103] The beneficial effect is that multi-scale coarse-grained analysis of the pressure data sequence after cleaning is performed. By dividing the data into fixed-length segments and calculating the mean, a coarse-grained sequence is obtained, which simplifies the data complexity and retains the core features of pressure changes at different scales, laying the foundation for subsequent accurate extraction of leakage-related signal patterns.

[0104] By topologically reconstructing coarse-grained sequences to construct time-delayed phase spaces, one-dimensional data is transformed into high-dimensional spatial vectors, clearly presenting the temporal correlation and evolution patterns between data. This breaks the limitation that single-dimensional data is difficult to capture complex pressure changes, and makes the latent signal characteristics caused by leakage explicit.

[0105] By identifying arrangement patterns and statistically analyzing their distribution probabilities in the time-delayed phase space, the inherent arrangement rules of coarse-grained sequences can be accurately captured. The frequency of occurrence of different arrangement patterns can be quantified, providing a comprehensive probabilistic basis for subsequent entropy calculation and ensuring the targeting and accuracy of feature extraction.

[0106] Information entropy quantification of the probability distribution yields multi-scale permutation entropy, transforming the probability distribution into a quantitative indicator that can characterize the complexity of the data. This indicator can accurately distinguish between normal pressure signals and abnormal signals caused by leakage, providing core complexity feature support for high-dimensional signal fingerprint feature vectors and improving the accuracy of leakage identification.

[0107] The pressure data sequence after cleaning is transformed by time-frequency transformation, converting the single time domain data into a time-frequency signal representation in the joint time-frequency domain. This fully presents the intensity distribution of the pressure signal at different times and frequencies, captures the time-frequency characteristic changes caused by leakage, and provides a comprehensive data foundation for energy characterization.

[0108] Energy values ​​are used to characterize the time-frequency representation of the signal. By calculating the squared value of the signal amplitude at each time-frequency coordinate point, the energy information under different spatiotemporal combinations is quantified to form an instantaneous energy spectrum, which accurately reflects the energy fluctuation differences caused by leakage and provides core energy features for feature fusion.

[0109] By vectorizing and concatenating the instantaneous energy spectrum with the multi-scale permutation entropy, the energy features and data complexity features are organically combined to obtain a preliminary fingerprint feature vector. This breaks through the limitations of a single feature dimension, comprehensively covers the signal features related to leakage, and improves the completeness of feature representation.

[0110] The preliminary fingerprint feature vectors are normalized and integrated. By uniformly scaling, the differences in dimensions and numerical ranges of different types of features are eliminated, so that each feature component is in the same comparable range. This ensures the stability and consistency of the feature vectors and provides accurate and reliable high-dimensional feature support for subsequent construction of causal attribute maps and loss localization.

[0111] S3. Map the topological data of the water supply network to a directed weighted graph of the water supply network, and inject the high-dimensional signal fingerprint feature vector into the directed weighted graph to obtain the causal attribute graph of the water supply network;

[0112] In this embodiment of the invention, reconstructing the topology data of the water supply network into a directed weighted graph of the water supply network includes:

[0113] Collect the topology data of the water supply network;

[0114] Using the topology data as a reference, the connection relationships of the network components representing the water supply network are determined;

[0115] Based on the aforementioned characterization of the network component connection relationship, the network components of the topology data are defined as node groups;

[0116] Based on the connection relationship of the network components, a directed connection is made between the node groups to obtain the directed configuration connection group of the water supply network.

[0117] Based on the pipe segment attributes in the topology data, the directed connection groups are assigned corresponding connection weights.

[0118] Based on the connection weights, the node groups are sequentially connected according to the directed connection groups to construct a directed weighted graph of the water supply network.

[0119] The step of injecting the high-dimensional signal fingerprint feature vector into the directed weighted graph to obtain the causal attribute graph of the water supply network includes:

[0120] The node location information in the topology data is spatially aligned with the acquisition location information of the real-time pressure monitoring data to obtain the spatial mapping relationship of the water supply network.

[0121] The time synchronization relationship of the water supply network is obtained by calibrating the timestamp of the real-time pressure monitoring data with the generation time of the high-dimensional signal fingerprint feature vector.

[0122] Based on the spatial mapping relationship and the time synchronization relationship, a multi-dimensional correlation mapping analysis is performed on the high-dimensional signal fingerprint feature vector and the corresponding monitoring node in the directed weighted graph to obtain the accurate matching relationship of the water supply network.

[0123] Based on the precise matching relationship, the high-dimensional signal fingerprint feature vector is used as an attribute feature and assigned to the corresponding node in the directed weighted graph to obtain the causal attribute graph of the water supply network.

[0124] Through various channels such as the water supply network surveying and mapping data acquisition system, equipment ledger records, and on-site survey records, we comprehensively collect topological data including information such as network component types, installation locations, pipe segment distribution, and connection ports to ensure that the collected data can completely and accurately reflect the actual layout and composition of the water supply network.

[0125] Using the collected topology data as the core reference benchmark, the physical connection relationships of various components in the pipeline network are sorted out one by one, the other components connected to the input and output ports of each component are clarified, the connection methods and associated paths between different pipeline network components are clearly defined, and finally, the connection relationship of pipeline network components that can accurately reflect the connection logic of water supply pipeline network components is determined.

[0126] Based on the established connection relationships of pipeline network components, all pipeline network components in the topology data are classified and grouped according to fixed classification rules such as component functional attributes, installation area or connection level. Each group contains a set of pipeline network components that are functionally related or have direct connection relationships. These groups are defined as node groups, so that the scattered pipeline network components form structured node units.

[0127] Based on the component connection logic defined by the connection relationship of the pipeline network components, directed connections are established between each node group. The connection direction strictly follows the actual flow direction of water in the pipeline network or the signal transmission direction between components, ensuring that each directed connection can accurately correspond to the actual interaction relationship between the node groups. In this way, directed connection groups of the water supply network are formed.

[0128] Key attribute information of pipe segments is extracted from the topology data, including pipe diameter, length, material, flow resistance coefficient, design flow rate, and other attribute parameters related to the network operation status. According to the preset weighting rules, these attribute parameters are converted into corresponding numerical weights, and each directed connection in the directed connection group is assigned a matching connection weight, so that the weight can quantitatively reflect the actual characteristics of the pipe segment.

[0129] Based on the assigned connection weights, and in accordance with the connection relationships and directions defined by the directed connection groups, all node groups are connected in an orderly sequence to ensure that the connections between node groups not only conform to the connection relationships of network components, but also reflect the differences in pipe segment attributes through connection weights. Finally, a directed weighted graph that can fully reflect the topology layout, connection direction and pipe segment characteristics of the water supply network is constructed.

[0130] The physical location description information of all nodes in the topology data is extracted, and the location description information of the collection points corresponding to the real-time pressure monitoring data is collected. The two types of location information are normalized using a unified spatial positioning standard. By comparing the spatial coordinate association, regional affiliation, and adjacent location references between the node location and the collection location one by one, it is ensured that each monitoring data collection location can accurately correspond to a specific node in the topology, forming a spatial mapping relationship that clearly defines the spatial correspondence between the two.

[0131] The system acquires the timestamp information of real-time pressure monitoring data generation and the generation time information of the high-dimensional signal fingerprint feature vector after data processing. Both types of time information are calibrated using the network's unified time as a benchmark. The format and accuracy of the time records are checked and adjusted to eliminate time differences caused by equipment clock deviations or data transmission delays, ensuring that the timestamp of each real-time pressure monitoring data point completely corresponds to the generation time of the corresponding high-dimensional signal fingerprint feature vector, thus obtaining the time synchronization relationship of the water supply network.

[0132] Based on spatial mapping relationships, the directed weighted graph nodes associated with the monitoring locations corresponding to the high-dimensional signal fingerprint feature vectors are determined. Combined with temporal synchronization relationships, feature vectors and node data within the same time dimension are locked. A comprehensive analysis is conducted on the high-dimensional signal fingerprint feature vectors and corresponding monitoring nodes in the directed weighted graph from multiple dimensions, including spatial correspondence, temporal matching, and data association logic, ensuring complete consistency between feature vectors and nodes in space, time, and data content, thus obtaining a precise matching relationship for the water supply network.

[0133] Based on precise matching relationships, the specific nodes in the directed weighted graph corresponding to each high-dimensional signal fingerprint feature vector are identified. The time-frequency energy information, sequence complexity, and other attribute features contained in the high-dimensional signal fingerprint feature vector are fully assigned to the corresponding nodes. Feature vector-related fields are added to the node attribute system of the directed weighted graph, allowing nodes to retain their original topological attributes and connection weights while also adding pressure signal-related feature attributes. This ultimately forms a causal attribute graph that integrates topological structure, connection relationships, pipe segment characteristics, and pressure signal features.

[0134] The beneficial effects are that it comprehensively collects the topological structure data of the water supply network, and fully obtains core information such as the type of network components, installation location, and distribution of pipe segments, providing detailed and accurate basic data for subsequent map construction, and ensuring that the directed weighted map can truly reflect the actual layout of the network.

[0135] Based on topological data, the connection relationships of network components are determined, the connection logic and associated paths of various components are clearly defined, and deviations in map construction due to ambiguity in connection relationships are avoided, providing a clear basis for node group division and directed connection.

[0136] Based on the characterization of the connection relationship of pipeline network components, the pipeline network components are defined as node groups, which makes the dispersed pipeline network components form structured units, simplifies the analysis dimension of complex pipeline networks, strengthens the functional correlation between components, and lays a regular foundation for subsequent directed connections.

[0137] Based on the connection relationship of the network components, directed connections are established between node groups. The connection direction is consistent with the actual flow direction of water flow or signal transmission in the network, forming a directed connection group. This accurately restores the transfer logic of matter and energy in the network and improves the practicality of the map.

[0138] By combining the pipe segment attributes in the topology data, the directed connection groups are assigned connection weights, which quantify the key characteristics of the pipe segments such as pipe diameter and material. This allows the directed weighted graph to not only reflect the connection relationships but also the physical differences in the operation of the pipeline network, providing quantitative support for subsequent anomaly prediction.

[0139] Based on the connection weights and the directed connection groups of the node groups, a complete directed weighted graph is constructed, realizing the organic integration of pipeline network topology, connection direction and pipe segment characteristics. This provides a structured and visualized analysis carrier for high-dimensional signal fingerprint feature vector injection and pressure anomaly inference, significantly improving the efficiency and accuracy of leakage location.

[0140] By aligning the node positions of the topology data with the acquisition positions of the real-time pressure monitoring data through spatial consistency, a spatial mapping relationship is established to ensure that the pressure monitoring data corresponds accurately to the physical nodes of the pipeline network. This avoids deviations in feature vector injection due to spatial misalignment and lays the foundation for spatial matching in subsequent correlation analysis.

[0141] The timestamps of real-time pressure monitoring data and the generation times of high-dimensional signal fingerprint feature vectors are time-synchronized and calibrated to obtain a time synchronization relationship, eliminate deviations in the time dimension, and ensure that the association between feature vectors and nodes is based on the pressure state in the same time dimension, thereby improving the timeliness and accuracy of data association.

[0142] Multi-dimensional correlation mapping analysis is carried out based on spatial mapping relationship and temporal synchronization relationship. The high-dimensional signal fingerprint feature vector and the monitoring node in the directed weighted graph are fully matched from the spatial, temporal and data logic levels to obtain accurate matching relationship and ensure that the binding of feature vector and corresponding node is complete and without mismatch.

[0143] Based on the precise matching relationship, the high-dimensional signal fingerprint feature vector is assigned to the corresponding node, so that the directed weighted graph retains the pipeline topology, connection direction and pipe segment weight information, and incorporates the high-dimensional feature attributes of the pressure signal, forming a causal attribute graph. This achieves a deep integration of pipeline structure and pressure anomaly characteristics, providing a structured and panoramic analysis carrier for subsequent pressure anomaly inference.

[0144] S4. Perform pressure anomaly deduction on the causal attribute map to obtain the potential propagation path and root cause node set of the water supply network;

[0145] In this embodiment of the invention, the step of performing pressure anomaly deduction on the causal attribute map to obtain the potential propagation path and root cause node set of the water supply network includes:

[0146] Extract the node attributes of the high-dimensional signal fingerprint feature vector from the nodes of the causal attribute graph;

[0147] Based on the node attributes, assess the degree of pressure state anomaly of the nodes in the causal attribute graph;

[0148] Based on the degree of pressure anomaly, the directed propagation process between the nodes is simulated to obtain the root node set of the water supply network.

[0149] By tracing the state propagation beginning of the root node, the resulting state propagation link is determined as the potential propagation path of the water supply network.

[0150] Complete attribute data for each node in the causal attribute graph is retrieved one by one. High-dimensional signal fingerprint feature vector related attributes, which contain core information such as time-frequency energy and sequence complexity, are accurately selected. These attributes are extracted and organized into independent node attribute datasets to ensure that the extracted node attributes can fully reflect the pressure signal characteristics corresponding to each node.

[0151] A fixed pressure status assessment standard is established, which is based on the pressure signal characteristics of the water supply network during normal operation and clarifies the boundary conditions between normal and abnormal conditions. The node attributes of each extracted node are compared with the assessment standard one by one. The deviation of each characteristic indicator in the node attributes from the standard value is analyzed. Based on the direction, range and severity of the deviation, the degree of pressure status abnormality of each node is quantitatively determined to ensure that the assessment results are objective and consistent.

[0152] Based on the degree of pressure anomaly at each node, and combined with the directed connections and connection weights between nodes in the causal attribute diagram, the propagation process of pressure anomalies between nodes is simulated. Following the direction of the directed connections and adhering to the propagation resistance or efficiency characteristics reflected by the connection weights, the diffusion path of the anomaly from nodes with high anomalies to nodes with low anomalies is traced. By repeatedly verifying the rationality of the propagation logic, the core nodes that initially triggered the propagation of the anomaly are identified, and these core nodes are integrated to form the root node set of the water supply network.

[0153] For each root node in the root node set, the initial propagation point of its abnormal pressure state is traced back. Following the directed connection links between nodes in the causal attribute diagram, the entire process of the abnormal state propagating from the root node through various related nodes is fully tracked. All nodes, connections, and propagation order involved in the propagation process are recorded, and this complete abnormal state propagation link is clearly defined as the potential propagation path of the water supply network, ensuring that the path accurately reflects the actual trajectory of the abnormal propagation.

[0154] The beneficial effects are that it can accurately extract the node attributes corresponding to the high-dimensional signal fingerprint feature vector from the nodes of the causal attribute graph, focus on the core feature data related to leakage, avoid interference from irrelevant attributes, and provide accurate and targeted analytical basis for the assessment of abnormal pressure status.

[0155] By comparing node attributes with preset normal pressure signal characteristic standards, the degree of pressure state abnormality of each node is quantitatively evaluated to ensure that the abnormality judgment is objective and consistent, accurately distinguish between normal nodes and abnormal nodes, and clarify the core objects for subsequent abnormality propagation simulation.

[0156] Based on the degree of abnormality in pressure state, the abnormal propagation process is simulated by combining the directed connection relationship and connection weight between nodes. The abnormal propagation path is accurately tracked, the root node that initially caused the abnormality is identified and a root node set is formed, avoiding blind investigation and improving the efficiency of tracing the source of loss.

[0157] By tracing the state propagation beginning of the root node, the entire process of the anomaly spreading from the root node to related nodes is fully reconstructed, forming a clear potential propagation path, clarifying the context and key nodes of the anomaly propagation, and providing clear path guidance for subsequent reverse traversal to locate the leaked interval.

[0158] S5. Using the nodes in the root node set as indexes, traverse the associated branches of the potential propagation path in reverse, and combine the similarity clustering results of the associated branches to accurately locate the leakage range of the water supply network.

[0159] In this embodiment of the invention, the step of using nodes in the root source node set as indexes to traverse the associated branches of the potential propagation path in reverse, and combining the similarity clustering results of the associated branches to accurately locate the leakage range of the water supply network includes:

[0160] Based on the degree of pressure anomaly of the root node on the potential propagation path, the reverse traversal priority for the potential propagation path is dynamically allocated;

[0161] Based on the reverse traversal priority, a progressive traversal is performed in the opposite direction of the connection in the directed weighted graph, and the topological connection structure and signal fingerprint features of the nodes passed through are aggregated in real time during the traversal process to obtain the candidate associated branches of the water supply network.

[0162] Based on the candidate association branches, construct the fusion similarity index of the water supply network;

[0163] Based on the fusion similarity index, all candidate association branches are grouped and clustered to obtain the clustering groups of the water supply network;

[0164] From the clustering groups, the group with the highest fusion similarity index and the most significant abnormal performance of the signal fingerprint features is identified as the core abnormal group of the water supply network;

[0165] Based on the core anomaly group, the continuous pipe sections of the water supply network are determined to obtain the leakage range of the water supply network.

[0166] The formula for calculating the fusion similarity index is as follows:

[0167] ;

[0168] In the formula, The fusion similarity index, The topological similarity components between different branches in the candidate association branches. The feature similarity components between different branches in the candidate association branches. This refers to the dynamic adjustment coefficient of the candidate association branch. Let A be the topological depth of branch A among the candidate related branches. Let B be the topological depth of the candidate associated branches. The branch normalization factor is the number of candidate related branches. It is an exponential function.

[0169] A thorough analysis of the anomaly level of each node in the root node set along potential propagation paths is conducted. Root nodes with more severe anomalies are assigned higher priority for reverse traversal, thus establishing a clear priority ranking rule. Following this rule, the order of reverse traversal is dynamically assigned to each root node, ensuring that traversal is prioritized for path branches with more significant anomaly impacts, thereby improving the targeting and efficiency of leaky interval location.

[0170] Strictly adhering to the established reverse traversal priority, the process begins with the root node of the highest priority and proceeds progressively along the reverse direction of the connections between nodes in the directed weighted graph. During the traversal, topological connection structure information of each node is collected in real time, including the node's connection method, the number of associated nodes, and pipe segment attributes. Simultaneously, signal fingerprint feature data of each node is aggregated. This information is then integrated and sorted to form candidate associated branches of the water supply network.

[0171] For each candidate association branch, the similarity features of its topological connection structure and the correlation attributes of its signal fingerprint features are extracted, and a fusion similarity index is constructed based on these features. This index comprehensively considers the fit of the topological structure between branches and the correlation of signal features, and can comprehensively reflect the similarity between different candidate association branches, providing a unified judgment basis for subsequent clustering analysis.

[0172] Using the fusion similarity index as the core criterion, all candidate association branches are grouped and clustered. Candidate association branches with high fusion similarity indices and similar feature performance are grouped into the same group to ensure that branches within the same cluster group have strong consistency in topology and signal features, and that there are significant differences between different groups, thus obtaining the clustering groups of the water supply network.

[0173] The fusion similarity index value and the abnormal signal fingerprint characteristics of the nodes within each cluster were analyzed one by one, and the index levels and the degree of anomaly significance among different groups were compared. The cluster with the highest fusion similarity index value and the most prominent abnormal signal fingerprint characteristics of the nodes within the group were selected and identified as the core anomaly group of the water supply network, focusing on key areas related to leakage.

[0174] Based on the nodes and connections within the core anomaly group, the actual pipe segment distribution corresponding to the nodes within the group is analyzed, and the continuous pipe segments associated with these nodes are identified. By confirming the physical continuity and correlation of the pipe segments, the interference of discrete nodes is eliminated, and the actual pipe network area corresponding to the core anomaly group is accurately located. This area is the leakage range of the water supply network.

[0175] The fusion similarity index is derived from the topological similarity component, feature similarity component, dynamic adjustment coefficient, topological depth, and branch normalization factor of the candidate related branches. It is generated by integrating these components through specific operational logic and is the core indicator for measuring the degree of similarity between candidate related branches.

[0176] The topological similarity component is derived from the topological connection structure of different branches in the candidate association branches. The topological information such as the connection method, node association relationship, and pipe segment distribution characteristics of different branches are compared and analyzed to extract their similarity features and convert them into corresponding quantitative values ​​to obtain the topological similarity component.

[0177] The feature similarity component is derived from the signal fingerprint features of different branches in the candidate association branch. By comparing and analyzing the node signal fingerprint features contained in different branches, including time-frequency energy information, sequence complexity, etc., the similarity at the feature level is extracted and quantified into feature similarity components.

[0178] The dynamic adjustment coefficient is set according to the actual situation of the candidate association branches and the clustering requirements. It is used to balance the weight ratio of topological similarity component and feature similarity component in the fusion similarity index, so as to ensure that the index can adapt to the similarity judgment requirements in different scenarios.

[0179] The topology depth of branch A is derived from branch A in the candidate associated branches. Based on the topology information such as the node hierarchy distribution and connection path length of this branch, its depth attribute in the entire pipeline topology is determined and quantified.

[0180] The topology depth of branch B is derived from branch B in the candidate associated branches. Using the same calculation method as the topology depth of branch A, the hierarchical and path characteristics in its topology structure are analyzed, and the corresponding topology depth value is quantified.

[0181] The branch normalization factor is determined based on the topological depth distribution range of all candidate related branches. It is used to standardize the differences in topological depth between different branches and avoid the accuracy of similarity determination being affected by excessive differences in topological depth values.

[0182] The exponential function is a pre-defined fixed operation rule used to transform the result after the topological depth difference has been normalized, so that the impact of this part on the fusion similarity index conforms to the pre-defined logical rules and improves the rationality of the index.

[0183] The significance of the formula lies in comprehensively considering the topological similarity and signal feature similarity between candidate related branches, balancing the weights of the two by dynamically adjusting the coefficients, and making corrections based on the differences in topological depth between different branches.

[0184] This operational logic can comprehensively and accurately quantify the overall similarity between candidate related branches, taking into account both the correlation at the topological level and the consistency at the signal feature level. It also avoids misjudgment of similarity caused by differences in branch structure levels through topological depth correction.

[0185] The final generated fusion similarity index provides a scientific and unified judgment standard for the grouping and clustering of candidate associated branches, ensuring that the clustering results can accurately reflect the actual degree of association between branches, and laying a reliable foundation for the subsequent accurate location of leakage intervals in the water supply network.

[0186] The beneficial effect is that the reverse traversal priority is allocated according to the degree of pressure anomaly of the root node on the potential propagation path. The more severe the anomaly, the higher the priority, ensuring that the core abnormal area is focused first, avoiding blind traversal, and improving the targeting and efficiency of loss location.

[0187] Based on priority, the system traverses the directed weighted graph in reverse direction, aggregates the node topology and signal fingerprint features in real time, forms candidate association branches, and fully collects the branch data related to leakage, providing comprehensive and accurate basic materials for subsequent clustering analysis.

[0188] Based on the candidate association branches, a fusion similarity index is constructed, which comprehensively considers the similarity of topological structure and signal features, providing a scientific and unified judgment standard for branch grouping and avoiding clustering bias caused by single-dimensional judgment.

[0189] Based on the fusion similarity index, candidate related branches are grouped and clustered, so that branches with similar characteristics are grouped into one category, forming cluster groups, clearly distinguishing normal branches from abnormal branches, and narrowing the scope of missed investigation.

[0190] By selecting the core anomaly group with the highest fusion similarity index and the most significant abnormal signal fingerprint features from the clustering group, the key branch set directly related to the leakage is accurately identified, providing the core basis for determining the leakage interval.

[0191] By identifying continuous pipe sections based on the core anomaly group, the specific range of the leakage area is clarified, solving the problems of vague range and low accuracy in traditional positioning, providing accurate location guidance for subsequent leakage repair, and significantly improving the efficiency of water supply network leakage management.

[0192] The formula integrates the topological similarity component and the feature similarity component, and balances the weight ratio of the two in the fusion similarity index by dynamically adjusting the coefficient. It not only attaches importance to the fit of the topological structure of the candidate related branches, but also takes into account the correlation of signal features, so as to ensure that the index can fully reflect the similar essence between branches.

[0193] The topological depth difference between branch A and branch B is introduced, and the branches are standardized by a branch normalization factor. Then, the exponential function transformation is combined to effectively correct the misjudgment of similarity caused by the difference in topological level between different branches, and improve the adaptability of the indicator to the branch structure characteristics.

[0194] The formula generates a quantitative fusion similarity index through multi-parameter collaborative operation, providing a unified and accurate judgment standard for the grouping and clustering of candidate related branches, avoiding the bias caused by subjective judgment or single-dimensional evaluation, and ensuring the objectivity and reliability of the clustering results.

[0195] The fusion similarity index obtained based on this formula can accurately distinguish between normal branches and abnormal branches, helping to quickly screen out the core abnormal groups related to leakage, providing key data support for the accurate location of subsequent leakage intervals, and significantly improving the efficiency and accuracy of leakage location in water supply networks.

[0196] S6. A comprehensive assessment of the leakage area is conducted to obtain a leakage location diagnosis report for the water supply network.

[0197] In this embodiment of the invention, a comprehensive analysis of the leakage range is performed to obtain a leakage location and diagnosis report for the water supply network.

[0198] Collect complete data corresponding to the leakage range, including topological connection details of nodes within the range, abnormal data of signal fingerprint features, pressure status monitoring records, pipe segment attribute parameters, etc. At the same time, summarize the relevant data in the process of potential propagation paths, root cause node sets and cluster analysis to ensure that the data required for the judgment is comprehensive and accurate.

[0199] The data within the leakage range are analyzed hierarchically, focusing on the anomaly types, intensities, and distribution patterns of signal fingerprint features. The critical path of anomaly propagation is determined by combining the topological connection structure. Pressure change trends that may be caused by leakage are verified through pressure state data, and the core anomaly manifestations and related logic within the leakage range are clarified.

[0200] By combining historical leakage cases of water supply networks, the service life of pipe sections, maintenance records, and other background information, the abnormal characteristics of the current leakage area are compared with the similarities of historical cases. The factors influencing the probability of leakage are assessed by referring to attributes such as pipe material and installation process, providing empirical support and practical basis for leakage diagnosis.

[0201] Professional and technical personnel were organized to conduct a comprehensive analysis of the analysis results and background information to clarify the specific range of the leakage area, the possible types of leakage, and the severity of leakage. At the same time, the potential causes of leakage were analyzed, including pipeline aging, external damage, and construction defects, to form a comprehensive analysis conclusion.

[0202] Following a standardized report format, the results of the leakage location, data analysis process, judgment basis, leakage type and cause analysis, and potential impact assessment are structured and organized to ensure that the report is logically clear, detailed, and accurate. This results in a leakage location and diagnosis report for the water supply network, providing clear guidance for subsequent leakage repair and network maintenance.

[0203] The beneficial effects are that the comprehensive assessment process collects core data such as the topological connectivity, signal characteristics, and pressure monitoring of the leakage area, while integrating the previous analysis results such as potential propagation paths and root cause node sets, ensuring that the assessment basis is detailed and sufficient, and providing comprehensive data support for the diagnostic report.

[0204] By analyzing data within the leakage range in layers, we can accurately extract the types, intensities, and propagation logic of abnormal signals. Combined with historical cases of the pipeline network and the attributes of the pipeline sections, we can conduct in-depth analysis of the leakage type, severity, and causes, making the diagnostic conclusions more scientific and convincing.

[0205] By combining data analysis and background information with professional personnel to conduct comprehensive assessments, the limitations of a single data dimension are avoided, enabling a holistic evaluation of leakage and ensuring that the diagnostic results are objective, accurate, and consistent with the actual operating conditions of the pipeline network.

[0206] The results of the analysis are organized in a standardized format to form a well-structured and detailed leakage location and diagnosis report, which clarifies the scope, cause and potential impact of the leakage, and provides clear guidance for the formulation of subsequent leakage repair plans and the optimization of pipeline maintenance, thus significantly improving the pertinence and efficiency of water supply network leakage management.

[0207] like Figure 2 The diagram shown is a functional block diagram of a water supply network leakage location system provided in an embodiment of the present invention.

[0208] The water supply network leakage location system 100 described in this invention can be installed in an electronic device. Depending on the functions implemented, the water supply network leakage location system 100 may include a standard cleaning module 101, a fusion vector module 102, an image construction module 103, an anomaly inference module 104, a leakage location module 105, and a location diagnosis module 106. The module described in this invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0209] In this embodiment, the functions of each module / unit are as follows:

[0210] The standard cleaning module 101 is used to perform adaptive filtering and robust outlier cleaning on the real-time pressure monitoring data of the water supply network to obtain the pressure data sequence of the water supply network after cleaning.

[0211] The fusion vector module 102 is used to fuse the multi-scale permutation entropy of the pressure data sequence after cleaning with the instantaneous energy spectrum into a high-dimensional signal fingerprint feature vector of the water supply network.

[0212] The image construction module 103 is used to reconstruct the topological structure data of the water supply network into a directed weighted graph of the water supply network, and inject the high-dimensional signal fingerprint feature vector into the directed weighted graph to obtain the causal attribute graph of the water supply network.

[0213] The anomaly inference module 104 is used to perform pressure anomaly inference on the causal attribute map to obtain the potential propagation path and root node set of the water supply network.

[0214] The leakage location module 105 is used to traverse the associated branches of the potential propagation path in reverse using the nodes in the root source node set as indexes, and combine the similarity clustering results of the associated branches to accurately locate the leakage range of the water supply network.

[0215] The location diagnosis module 106 is used to comprehensively analyze the leakage range and obtain a leakage location diagnosis report of the water supply network.

[0216] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0217] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0218] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0219] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0220] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0221] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for locating leakage in a water supply network, characterized in that, The method includes: S1. Adaptive filtering and robust outlier cleaning are performed on the real-time pressure monitoring data of the water supply network to obtain the cleaned pressure data sequence of the water supply network; S2. The multi-scale permutation entropy of the pressure data sequence after cleaning and the instantaneous energy spectrum are fused into a high-dimensional signal fingerprint feature vector of the water supply network; S3. Map the topological data of the water supply network to a directed weighted graph of the water supply network, and inject the high-dimensional signal fingerprint feature vector into the directed weighted graph to obtain the causal attribute graph of the water supply network; S4. Perform pressure anomaly deduction on the causal attribute map to obtain the potential propagation path and root cause node set of the water supply network; S5. Using the nodes in the root node set as indexes, traverse the associated branches of the potential propagation path in reverse, and combine the similarity clustering results of the associated branches to accurately locate the leakage range of the water supply network. S6. A comprehensive assessment of the leakage area is conducted to obtain a leakage location diagnosis report for the water supply network.

2. The method for locating leakage in a water supply network as described in claim 1, characterized in that, The real-time pressure monitoring data of the water supply network is subjected to adaptive filtering and robust outlier cleaning to obtain a post-cleaning pressure data sequence of the water supply network, including: Receive real-time pressure monitoring data of the water supply network to obtain the original pressure sequence of the water supply network; The original pressure sequence is adaptively filtered to obtain the filtered pressure sequence of the water supply network. Outlier identification is performed on the filtered pressure sequence to obtain the anomaly marker sequence of the water supply network; Based on the anomaly marker sequence, the filtered pressure sequence is cleaned to obtain the cleaned pressure data sequence of the water supply network.

3. The method for locating leakage in a water supply network as described in claim 1, characterized in that, The process of fusing the multi-scale permutation entropy of the post-cleaning pressure data sequence with the instantaneous energy spectrum to form a high-dimensional signal fingerprint feature vector for the water supply network includes: Multi-scale coarse-grained analysis was performed on the pressure data sequence after cleaning to obtain the coarse-grained sequence of the water supply network; The coarse-grained sequence is topologically reconstructed to obtain the time-delay phase space of the water supply network; In the time-delay phase space, the arrangement pattern of the coarse-grained sequence is identified to statistically determine the distribution probability of the coarse-grained sequence; The information entropy quantization process is performed on the probability distribution to obtain the multi-scale arrangement entropy of the water supply network.

4. The method for locating leakage in a water supply network as described in claim 3, characterized in that, The process of fusing the multi-scale permutation entropy of the post-cleaning pressure data sequence with the instantaneous energy spectrum to form a high-dimensional signal fingerprint feature vector for the water supply network includes: The time-frequency representation of the pressure data sequence after cleaning is obtained by performing a time-frequency transformation on the pressure data sequence after cleaning. The instantaneous energy spectrum of the water supply network is obtained by characterizing the energy value of the time-frequency representation of the signal. The instantaneous energy spectrum and the multi-scale arrangement entropy are vectorized and concatenated to obtain the preliminary fingerprint feature vector of the water supply network; The preliminary fingerprint feature vector is normalized and integrated to obtain the high-dimensional signal fingerprint feature vector of the water supply network.

5. The method for locating leakage in a water supply network as described in claim 1, characterized in that, The step of reconstructing the topology data of the water supply network into a directed weighted graph of the water supply network includes: Collect the topology data of the water supply network; Using the topology data as a reference, the connection relationships of the network components representing the water supply network are determined; Based on the aforementioned characterization of the network component connection relationship, the network components of the topology data are defined as node groups; Based on the connection relationship of the network components, a directed connection is made between the node groups to obtain the directed configuration connection group of the water supply network. Based on the pipe segment attributes in the topology data, the directed connection groups are assigned corresponding connection weights. Based on the connection weights, the node groups are sequentially connected according to the directed connection groups to construct a directed weighted graph of the water supply network.

6. The method for locating leakage in a water supply network as described in claim 5, characterized in that, The step of injecting the high-dimensional signal fingerprint feature vector into the directed weighted graph to obtain the causal attribute graph of the water supply network includes: The node location information in the topology data is spatially aligned with the acquisition location information of the real-time pressure monitoring data to obtain the spatial mapping relationship of the water supply network. The time synchronization relationship of the water supply network is obtained by calibrating the timestamp of the real-time pressure monitoring data with the generation time of the high-dimensional signal fingerprint feature vector. Based on the spatial mapping relationship and the time synchronization relationship, a multi-dimensional correlation mapping analysis is performed on the high-dimensional signal fingerprint feature vector and the corresponding monitoring node in the directed weighted graph to obtain the accurate matching relationship of the water supply network. Based on the precise matching relationship, the high-dimensional signal fingerprint feature vector is used as an attribute feature and assigned to the corresponding node in the directed weighted graph to obtain the causal attribute graph of the water supply network.

7. The method for locating leakage in a water supply network as described in claim 1, characterized in that, The pressure anomaly extrapolation of the causal attribute map yields the potential propagation paths and root cause node set of the water supply network, including: Extract the node attributes of the high-dimensional signal fingerprint feature vector from the nodes of the causal attribute graph; Based on the node attributes, assess the degree of pressure state anomaly of the nodes in the causal attribute graph; Based on the degree of pressure anomaly, the directed propagation process between the nodes is simulated to obtain the root node set of the water supply network. By tracing the state propagation beginning of the root node, the resulting state propagation link is determined as the potential propagation path of the water supply network.

8. The method for locating leakage in a water supply network as described in claim 1, characterized in that, The step of using nodes in the root node set as indexes to traverse the associated branches of the potential propagation path in reverse, and combining the similarity clustering results of the associated branches to accurately locate the leakage range of the water supply network includes: Based on the degree of pressure anomaly of the root node on the potential propagation path, the reverse traversal priority for the potential propagation path is dynamically allocated; Based on the reverse traversal priority, a progressive traversal is performed in the opposite direction of the connection in the directed weighted graph, and the topological connection structure and signal fingerprint features of the nodes passed through are aggregated in real time during the traversal process to obtain the candidate associated branches of the water supply network. Based on the candidate association branches, construct the fusion similarity index of the water supply network; Based on the fusion similarity index, all candidate association branches are grouped and clustered to obtain the clustering groups of the water supply network; From the clustering groups, the group with the highest fusion similarity index and the most significant abnormal performance of the signal fingerprint features is identified as the core abnormal group of the water supply network; Based on the core anomaly group, the continuous pipe sections of the water supply network are determined to obtain the leakage range of the water supply network.

9. The method for locating leakage in a water supply network as described in claim 8, characterized in that, The formula for calculating the fusion similarity index is as follows: ； In the formula, The fusion similarity index, The topological similarity components between different branches in the candidate association branches. The feature similarity components between different branches in the candidate association branches. This refers to the dynamic adjustment coefficient of the candidate association branch. Let A be the topological depth of branch A among the candidate related branches. Let B be the topological depth of the candidate associated branches. The branch normalization factor is the number of candidate related branches. It is an exponential function.

10. A water supply network leakage location system, characterized in that, The system for implementing the water supply network leakage location method according to claim 1 includes: A standard cleaning module is used to perform adaptive filtering and robust outlier cleaning on real-time pressure monitoring data of the water supply network to obtain a pressure data sequence of the water supply network after cleaning. The fusion vector module is used to fuse the multi-scale permutation entropy of the pressure data sequence after cleaning with the instantaneous energy spectrum into a high-dimensional signal fingerprint feature vector of the water supply network. An image construction module is used to reconstruct the topological data of the water supply network into a directed weighted graph of the water supply network, and inject the high-dimensional signal fingerprint feature vector into the directed weighted graph to obtain the causal attribute graph of the water supply network. Anomaly simulation module is used to perform pressure anomaly simulation on the causal attribute map to obtain the potential propagation path and root cause node set of the water supply network. The leakage location module is used to traverse the associated branches of the potential propagation path in reverse using the nodes in the root node set as indexes, and combine the similarity clustering results of the associated branches to accurately locate the leakage range of the water supply network. The location diagnosis module is used to comprehensively analyze the leakage range and obtain a leakage location diagnosis report for the water supply network.

Images