A gas extraction pipe network leakage diagnosis positioning method based on branch pressure difference characteristics and topology constraints

Abstract

The application discloses a gas extraction pipe network leakage diagnosis and positioning method based on branch pressure difference characteristics and topological constraints. According to the topological mapping relationship among the nodes, pipe sections, pressure measuring points and candidate leakage points of the gas extraction pipe network, the time alignment, missing compensation, abnormality rejection, unit unification and smoothing processing are carried out on the multi-measuring-point pressure time series data. The adaptive pressure baseline of each pressure measuring point is constructed based on the stable operation period, the pressure offset is calculated and the multi-measuring-point joint abnormality statistics is constructed to determine the leakage abnormality interval. The branch pressure difference characteristics and multi-measuring-point pressure response characteristics between the upstream and downstream paired pressure measuring points of the candidate leakage points in the leakage abnormality interval are extracted, the candidate leakage position set is screened in combination with the topological constraint scoring, and the leakage start and end time, the most possible leakage position and the confidence information are output through an auxiliary discrimination model. The application can improve the stability and positioning accuracy of the gas extraction pipe network leakage identification, reduce the confusion between adjacent leakage points or pressure response similar leakage points, and realize the collaborative performance of leakage alarm, inspection prompt and maintenance positioning.

Description

[0001] This invention belongs to the field of intelligent monitoring and fault diagnosis technology for coal mine gas extraction, specifically involving a method for diagnosing and locating leaks in gas extraction pipelines based on branch pressure difference characteristics and topological constraints. Background Technology

[0002] Gas drainage pipelines are a crucial infrastructure for gas control and safe production in coal mines. Their operational status directly affects drainage efficiency, drainage concentration, negative pressure distribution, and mine safety. During the drainage process, the pipeline network is prone to leakage failures due to factors such as sealing failure of connecting components, aging of local pipe sections, fluctuations in valve operating conditions, and disturbances in branch lines.

[0003] Existing methods for identifying leaks in gas drainage pipeline networks mostly rely on manual inspections, fixed threshold alarms, or single data-driven models. Manual inspections have poor real-time performance; fixed threshold methods are prone to false alarms and missed alarms when boundary conditions change or when the baseline of the measuring points is inconsistent; single supervised learning methods are prone to confusing the location results when there are few samples and the responses of adjacent leak points are similar; in addition, gas drainage pipeline networks have obvious topological characteristics, and after a leak occurs at a certain location, its impact is not only reflected in the measuring points near the leak point, but also propagates and attenuates along the main pipe and adjacent branches.

[0004] Therefore, this invention proposes a leak diagnosis and location method that integrates multi-point pressure timing, adaptive stable baseline, multi-point joint anomaly detection, leakage pressure characteristics and topological constraint relationship, so as to improve the detection accuracy, location accuracy and engineering applicability in complex extraction pipeline network scenarios. Summary of the Invention

[0005] The purpose of this invention is to provide a method for diagnosing and locating leaks in gas extraction pipelines based on branch pressure difference characteristics and topological constraints, addressing the shortcomings of existing technologies.

[0006] To achieve the above objectives, this invention provides a method for diagnosing and locating leaks in gas extraction pipeline networks based on branch pressure difference characteristics and topological constraints, specifically including the following steps:

[0007] Step S1: Receive pressure time series data from multiple pressure measuring points in the gas extraction pipeline network, as well as topology mapping information between nodes, pipe sections, pressure measuring points, and candidate leak points.

[0008] Furthermore, the pressure time-series data can originate from a real-time on-site acquisition system, a historical database of an industrial monitoring platform, a host computer acquisition system, or imported data from external files. The topology mapping information is used to characterize the connection relationships between nodes and pipe segments in the pipeline network, the correspondence between pressure measuring points and pipe segments or nodes, the correspondence between candidate leak points and their upstream and downstream pressure measuring points, and the adjacency relationships or flow path relationships between candidate leak points.

[0009] Step S2: Preprocess the pressure time series data to obtain a multi-point pressure sequence that meets the requirements for leak diagnosis and location.

[0010] Furthermore, the preprocessing includes one or more of the following: time alignment, missing data compensation, anomaly removal, unit unification, and smoothing. Preprocessing enables pressure data from different measuring points, sources, or sampling intervals to be compared and calculated on a unified time axis, and reduces the impact of short-term noise and anomalous jumps on leak diagnosis results.

[0011] Step S3: Construct an adaptive pressure baseline for each pressure measurement point based on the stable operating period, and construct a joint anomaly statistic for multiple measurement points based on the pressure offset to determine the leakage anomaly range.

[0012] Furthermore, in step S3, firstly, a stable operating period is identified based on the average pressure, fluctuation level, and change amount of adjacent windows within the sliding time window; then, an adaptive pressure baseline for each pressure measuring point under the current operating condition is constructed using the data from the stable operating period; next, the pressure offset of each pressure measuring point relative to the adaptive pressure baseline is calculated, and the start and end times of leakage anomaly are determined based on the joint abnormal response of multiple pressure measuring points.

[0013] Step S4: Within the abnormal leakage range, extract the branch pressure difference characteristics between paired pressure measuring points upstream and downstream of the candidate leakage point, as well as the pressure response characteristics of multiple measuring points, to construct leakage pressure characteristics.

[0014] Furthermore, the branch pressure differential characteristic is used to characterize the local upstream and downstream pressure response differences at candidate leak points, and the multi-point pressure response characteristic is used to characterize the overall propagation characteristics of leakage disturbances in the pipeline network. By combining the branch pressure differential characteristic with the multi-point pressure response characteristic, a leakage pressure characteristic that can reflect the pressure response differences at different candidate leak locations is formed.

[0015] Step S5: Based on the topology mapping information, branch pressure difference characteristics and multi-measuring point pressure response characteristics, perform topology constraint scoring on each candidate leak point, and filter the candidate leak location set according to the scoring results.

[0016] Furthermore, the topology constraint score is determined based on the correlation between candidate leak points and adjacent pressure measuring points, the pressure difference change relationship between paired pressure measuring points upstream and downstream of the candidate leak point, and the attenuation relationship of the leak disturbance as it propagates along the pipeline network. By scoring and ranking each candidate leak point, one or more locations with the highest scores are selected as the set of candidate leak locations.

[0017] Step S6: Perform auxiliary discrimination on the candidate leak location set and output the leak diagnosis and location results.

[0018] Furthermore, the auxiliary discrimination is achieved by combining the topological constraint scoring results with the ECOC multi-class discrimination model based on support vector machines. The topological constraint scoring results are used to form a set of candidate leak locations, and the ECOC multi-class discrimination model based on support vector machines is used to further distinguish candidate leak locations that are topologically adjacent, close in distance, or have similar pressure responses, and output the most likely leak location.

[0019] Furthermore, the leak diagnosis and location results include the leak start time, leak end time, a set of candidate leak locations, the most likely leak location, and confidence information; when no abnormal leak interval is identified, a no-leak status is output. The confidence information is determined by one or more of the following: topological constraint score, auxiliary discrimination model output score, score difference between candidate leak locations, and multi-window voting results.

[0020] Compared with the prior art, the present invention has at least the following beneficial effects:

[0021] This invention constructs an adaptive pressure baseline based on pressure time-series data from multiple pressure measurement points, and determines the leakage anomaly range by pressure offset and joint anomaly statistics from multiple measurement points. This avoids the problem that the judgment based on a fixed threshold at a single measurement point is easily affected by operating condition fluctuations, measurement point differences, and measurement noise, thus improving the stability and reliability of leakage status identification.

[0022] This invention constructs branch pressure difference characteristics based on paired pressure measuring points upstream and downstream of candidate leak points, and combines multi-measuring point pressure response characteristics and pipeline topology mapping information to form leak pressure characteristics. It performs topology constraint scoring and auxiliary discrimination on candidate leak locations, so that the leak diagnosis results can correspond to specific candidate leak locations, which facilitates on-site inspection, maintenance location and subsequent control, and provides more sufficient feature basis for leak location judgment. Attached Figure Description

[0023] To make the objectives, methods and advantages of the present invention clearer, the present invention will now be described in further detail with reference to the accompanying drawings.

[0024] Figure 1 This is a flowchart illustrating a method for diagnosing and locating leaks in gas extraction pipelines based on branch pressure difference characteristics and topological constraints.

[0025] Figure 2 This is a schematic diagram illustrating the steps of a method for diagnosing and locating leaks in a gas extraction pipeline network based on branch pressure difference characteristics and topological constraints.

[0026] Figure 3 This is a schematic diagram showing the arrangement of gas extraction pipeline nodes, pipe sections, pressure measuring points, and candidate leak points in a specific embodiment of the present invention. Detailed Implementation

[0027] The present invention will now be described in detail with reference to the accompanying drawings and specific processes. Obviously, the following content is only used to illustrate the technical solution of the present invention and is not intended to limit the scope of protection of the present invention. Equivalent substitutions or modifications made by those skilled in the art based on the present invention without departing from the concept of the present invention should all fall within the scope of protection of the present invention.

[0028] like Figure 1 As shown, this invention provides a method for diagnosing and locating leaks in gas extraction pipeline networks based on branch pressure difference characteristics and topological constraints, specifically including the following steps:

[0029] Step S1: Receive pressure time series data from multiple pressure measuring points in the gas extraction pipeline network, as well as topology mapping information between nodes, pipe sections, pressure measuring points and candidate leak points.

[0030] like Figure 3 As shown in this specific embodiment, the gas extraction pipeline network includes nodes 1 to 14, pipe segments a to m, pressure measuring points P1 to P17, and candidate leak points F1 to F8. Nodes represent pipeline ends, branch points, or confluence points; pipe segments represent gas extraction connection pipelines between adjacent nodes; pressure measuring points collect pressure time-series data at corresponding locations; and candidate leak points represent potential leak locations pre-defined based on branch structure, connection points, and easily leaking locations on site.

[0031] Furthermore, such as Figure 3 As shown, node 14 is connected to node 13 via pipe segment m. Pressure measuring point P1 is arranged on pipe segment m to characterize the pressure response changes on the main pipe or the manifold side. Nodes 13, 12, 11, 10, and 9 are connected vertically in sequence, forming the central manifold connection structure of the network via pipe segments l, k, j, and i, respectively. Node 1 is connected to node 12 via pipe segment a, node 2 is connected to node 11 via pipe segment b, node 3 is connected to node 10 via pipe segment c, and node 4 is connected to node 9 via pipe segment d; node 12 is connected to node 5 via pipe segment e, node 11 is connected to node 6 via pipe segment f, node 10 is connected to node 7 via pipe segment g, and node 9 is connected to node 8 via pipe segment h. This forms a gas extraction pipeline network topology composed of the central manifold and multiple branches on both sides.

[0032] Furthermore, Figure 3 The gas extraction pipeline network shown is abstracted into a diagram:

[0033] Where V is the set of nodes and E is the set of pipe segments. Specifically, the set of nodes can be represented as:

[0034] Furthermore, the set of pipe segments can be represented as

[0035] Furthermore, the set of candidate leak points can be represented as

[0036] Furthermore, the set of pressure measurement points can be represented as

[0037] Furthermore, in Figure 3 In the specific embodiment shown, candidate leak points F1 to F8 are respectively set on eight branches, and pressure measuring points are arranged on both the upstream and downstream sides of each candidate leak point, forming a branch differential pressure monitoring relationship of single candidate leak point - dual pressure measuring points. Specifically, F1 is set on pipe segment a, corresponding to pressure measuring points P2 and P3; F2 is set on pipe segment b, corresponding to pressure measuring points P4 and P5; F3 is set on pipe segment c, corresponding to pressure measuring points P6 and P7; F4 is set on pipe segment d, corresponding to pressure measuring points P8 and P9; F8 is set on pipe segment e, corresponding to pressure measuring points P10 and P11; F7 is set on pipe segment f, corresponding to pressure measuring points P12 and P13; F6 is set on pipe segment g, corresponding to pressure measuring points P14 and P15; and F5 is set on pipe segment h, corresponding to pressure measuring points P16 and P17. Pressure measuring point P1 serves as a pressure measuring point on the main pipe or manifold side, used to assist in determining whether leakage disturbances propagate to the manifold and to enhance the stability of joint anomaly judgment by multiple measuring points.

[0038] Furthermore, let the multi-point pressure time-series vector composed of all pressure measuring points at time t be:

[0039] in, Indicates time Pressure time-series vector composed of each pressure measuring point; Indicates the first Pressure measuring point at time Pressure value, total number of pressure measuring points =17.

[0040] Step S2: Preprocess the pressure time series data to obtain a standardized multi-point pressure sequence that meets the requirements for leak diagnosis and location.

[0041] Furthermore, in this specific embodiment, the pressure time-series data collected from P1 to P17 undergo unified time axis alignment, missing value compensation, outlier removal, unit unification, and smoothing. For pressure measurement points with different sampling frequencies or short-term data gaps, resampling and interpolation methods are used for compensation; for jump points that significantly deviate from the pressure change pattern of adjacent time points, median filtering, moving average, or outlier threshold judgment methods are used for correction; for data from different sources, they are unified to the same pressure unit and the same time reference, thereby forming a standardized multi-point pressure sequence that can be used for subsequent leak anomaly detection and location analysis.

[0042] Step S3: Construct an adaptive pressure baseline for each pressure measuring point based on the stable operating period, calculate the pressure offset of each pressure measuring point relative to the adaptive pressure baseline, and construct a multi-measuring point joint anomaly statistic based on the pressure offset to determine the leakage anomaly range.

[0043] Furthermore, let the length of the sliding window be L, then the first... The window mean and standard deviation of the pressure measurement points in the k-th window are as follows:

[0044] Furthermore, when the conditions are met and The corresponding window is identified as a stable candidate window, where, For fluctuation threshold, This is the threshold for the change in the mean of adjacent windows.

[0045] Furthermore, in Figure 3 During the period when the pipeline network is operating normally and no leaks occur, a continuous time period with small pressure fluctuations and small changes in the mean value of adjacent windows is selected as the stable operating period. The set of stable candidate windows that continuously meet the stability judgment conditions and whose length reaches the preset requirement is denoted as . For the first For each pressure measurement point, its adaptive pressure baseline can be expressed as:

[0046] in, Indicates the first Adaptive pressure baseline for each pressure measurement point To stabilize the number of sample points in the candidate window set, an adaptive pressure baseline is constructed in the above manner, which can avoid misjudgments caused by changes in sampling negative pressure, differences in measurement point locations, and different static pressure levels when using a uniform fixed threshold or fixed reference value.

[0047] Furthermore, based on the aforementioned pressure baseline, the first... Pressure offset at each pressure measuring point at time t:

[0048] in, Indicates the first The pressure offset of each pressure measuring point relative to the stable operating baseline at time t. Figure 3 When a leak occurs at a candidate leak point, the upstream and downstream pressure measurement points near the leak point will first show a significant pressure shift. At the same time, the leak disturbance will propagate to the main pipe and adjacent branches through the confluence nodes such as node 9, node 10, node 11, node 12, and node 13, causing different degrees of pressure response at P1 and other branch measurement points.

[0049] Furthermore, to mitigate the impact of differences in static pressure levels and amplitude scales at different measuring points, the pressure offsets from P1 to P17 were standardized, and a multi-measuring-point joint anomaly statistic was constructed based on the continuous anomalies at multiple pressure measuring points. Let the first... The mean and standard deviation of the offset of each pressure measuring point are respectively and The corresponding outlier statistic is:

[0050] in, For the first Anomaly statistics of each pressure measuring point at time t It is a very small positive number, used to avoid the denominator being zero.

[0051] Furthermore, abnormal leakage periods are identified based on anomaly statistics from multiple pressure measurement points. Preferably, within a length of... Within the detection window, the first The percentage of pressure measurement points exceeding the threshold for an extended period is defined as:

[0052] in, For indicator functions, This is the threshold for outlier statistics.

[0053] Furthermore, the number of joint anomaly measurement points at time t is defined as:

[0054] When satisfied And the duration of this state is not less than And at least one key measurement point satisfies When the combined abnormal conditions are continuously not met and continue to be met, the leakage anomaly is determined to have started; When the leakage anomaly is determined to have ended, the leakage anomaly range is thus obtained. .in, For the threshold of the number of joint abnormal measurement points, For continuous proportion threshold, This is the threshold for the duration of the abnormal trigger. This is the time threshold for determining abnormal recovery.

[0055] Furthermore, through the above method, the present invention does not rely on a single instantaneous out-of-bounds signal at a single measuring point for alarm, but rather determines the leakage anomaly range by comprehensively considering the continuous abnormal conditions of multiple pressure measuring points, thereby improving the stability of identifying the start and end times of leakage.

[0056] Step S4: Within the abnormal leakage range, extract the branch pressure difference characteristics between paired pressure measuring points upstream and downstream of the candidate leakage point, as well as the pressure response characteristics of multiple measuring points, to construct leakage pressure characteristics.

[0057] Furthermore, within the aforementioned abnormal leakage range, according to Figure 3 The correspondence between the candidate leak points and the paired pressure measuring points upstream and downstream is shown, and eight sets of branch differential pressure characteristics are constructed respectively. Specifically, candidate leak point F1 corresponds to pressure measuring points P2 and P3, candidate leak point F2 corresponds to pressure measuring points P4 and P5, candidate leak point F3 corresponds to pressure measuring points P6 and P7, candidate leak point F4 corresponds to pressure measuring points P8 and P9, candidate leak point F8 corresponds to pressure measuring points P10 and P11, candidate leak point F7 corresponds to pressure measuring points P12 and P13, candidate leak point F6 corresponds to pressure measuring points P14 and P15, and candidate leak point F5 corresponds to pressure measuring points P16 and P17.

[0058] Specifically, at time The pressure difference characteristics of the eight branches are expressed as follows:

[0059] Furthermore, , , , , , , , These represent the branch pressure differences between the upstream and downstream paired pressure measuring points of candidate leak points F1, F2, F3, F4, F5, F6, F7, and F8, respectively.

[0060] Furthermore, to mitigate the impact of differences in initial pressure levels, measuring point installation locations, and normal extraction negative pressure on leakage assessment results across different branches, the change in differential pressure relative to the stable operating baseline for each group of branches was calculated. The change in differential pressure for the s-th group of branches is expressed as:

[0061] in, ; Indicates the first The pressure difference of the branch circuit at time The value; Indicates the first The baseline value of the differential pressure between the branch lines during stable operating periods; Indicates the first The pressure difference of the branch circuit at time The amount of change relative to a stable operating state.

[0062] Furthermore, the changes in pressure difference of the eight branches are expressed as follows:

[0063] Furthermore, in the abnormal leakage range Within this range, the average value of the differential pressure change in each branch group is calculated to characterize the overall differential pressure response intensity of the corresponding branch within the abnormal leakage range. The average variation characteristics of the branch pressure difference are expressed as follows:

[0064] Furthermore, The pressure difference change characteristics corresponding to candidate leak point F1, The pressure difference change characteristics corresponding to candidate leak point F2, The pressure difference change characteristics corresponding to candidate leak point F3, The pressure difference change characteristics corresponding to candidate leak point F4, The pressure difference change characteristics corresponding to candidate leak point F5, The pressure difference change characteristics corresponding to candidate leak point F6, The pressure difference change characteristics corresponding to candidate leak point F7, The pressure difference change characteristics corresponding to candidate leak point F8.

[0065] Furthermore, in the abnormal leakage range Furthermore, the multi-point pressure response characteristics of each pressure measuring point from P1 to P17 were extracted. The average pressure deviation characteristics of each pressure measuring point within the leakage anomaly range are expressed as follows:

[0066] in, Indicates the first The average pressure deviation characteristics of each pressure measuring point within the abnormal leakage range; Indicates the first Each pressure measuring point at time [time] Pressure offset relative to its adaptive pressure baseline; .

[0067] Furthermore, the leakage pressure characteristics are jointly constituted by eight sets of branch differential pressure change characteristics and pressure response characteristics of multiple measuring points P1 to P17. The leakage pressure characteristics include one or more of the following: average change in branch differential pressure, peak change in branch differential pressure, average offset of each pressure measuring point, peak offset of each pressure measuring point, dynamic response slope in the initial stage of the anomaly, differential pressure change amplitude ranking characteristics, and pressure response characteristics of the main pipe measuring point P1.

[0068] exist Figure 3 In the specific implementation shown, if a leak occurs at a candidate leak point, a significant pressure difference change will typically occur between the upstream and downstream paired pressure measuring points corresponding to that candidate leak point. For example, when a leak occurs near F1, the change in the branch pressure difference between P2 and P3 will be significant. The change in branch pressure difference between P8 and P9 is often more prominent when a leak occurs near F4. This is usually quite pronounced. Simultaneously, leakage disturbances propagate along the nodes and pipe segment connections to the main pipe and adjacent branches, causing varying degrees of pressure shifts at P1 and other branch measuring points. Therefore, this invention not only utilizes the pressure difference change of a single branch for judgment but also combines the pressure response characteristics of multiple measuring points to construct leakage pressure characteristics.

[0069] Step S5: Based on the topology mapping information, the branch pressure difference characteristics, and the multi-measuring point pressure response characteristics, perform topology constraint scoring on each candidate leak point, and filter the candidate leak location set according to the scoring results.

[0070] Furthermore, in Figure 3 In the gas extraction pipeline network shown, the system performs topology constraint scoring on candidate leak points F1, F2, F3, F4, F5, F6, F7, and F8. For the... Candidate leak points Construct the scoring function:

[0071] in, To match the degree of pressure difference variation in the branch, To determine the degree of matching of anomalous amplitudes at neighboring measuring points, For topology propagation attenuation consistency, These are the weighting coefficients.

[0072] Furthermore, candidate leak points The degree of matching of the corresponding branch pressure difference changes The leak point is determined based on the change in branch pressure difference between paired pressure measuring points upstream and downstream of the candidate leak point. When the corresponding branch pressure difference change is more prominent among the eight groups of branch pressure difference changes, it is considered that the branch pressure difference change of the candidate leak point has a high degree of matching.

[0073] Furthermore, candidate leak points degree of matching of abnormal amplitudes of adjacent pressure measuring points Based on the pressure deviation amplitude of the pressure measuring points near the candidate leak point within the abnormal leakage range, the nearby pressure measuring points are determined as follows: F1 is P2 and P3; F2 is P4 and P5; F3 is P6 and P7; F4 is P8 and P9; F5 is P16 and P17; F6 is P14 and P15; F7 is P12 and P13; and F8 is P10 and P11.

[0074] Furthermore, the consistency of the attenuation law of leakage disturbance propagation along the pipeline topology. Based on the topological connection between candidate leak points and other pressure measurement points, when a leak occurs at a candidate leak point, the pressure measurement points near the candidate leak point usually have a stronger response. As the leakage disturbance propagates along the pipe segment and node to the main pipe and adjacent branches, the response intensity of other measurement points gradually weakens. If the actual pressure response intensity of multiple measurement points is consistent with this topological propagation law, then the topological propagation attenuation consistency of the candidate leak point is high.

[0075] In this specific embodiment, if The corresponding branch shows the most significant pressure difference change, and the pressure offset amplitude near P2 and P3 is relatively large. Meanwhile, the pressure response at P1 and other branch measuring points follows a pattern of propagation from the branch where F1 is located to nodes 12, 13, and adjacent branches, gradually weakening. Therefore, candidate leak point F1 has a high topological constraint score, and the system considers F1 as a priority candidate leak location. If topologically adjacent locations such as F1, F8, and F2, or locations with similar pressure responses, have close scores, they are all included in the candidate leak location set.

[0076] Furthermore, based on the topological constraint scores of each candidate leak point F1 to F8, ranked from highest to lowest, one or more candidate leak points with the highest scores are selected to form a candidate leak location set, which can be represented as:

[0077] in, Represents the set of candidate leak locations; This indicates the candidate selection threshold. Alternatively, the candidate with the highest score can be selected based on the ranking results. The candidate leak points constitute the candidate leak location set.

[0078] Through the above topological constraint scoring process, the problem of locating leaks across the entire pipeline network can be transformed into a further discrimination problem within the set of candidate leak locations, thereby reducing confusion between adjacent leak points, symmetrical leak points, or leak points with similar pressure responses.

[0079] Step S6: Perform auxiliary discrimination on the candidate leak location set, and output the leak start time, leak end time, candidate leak location set, most likely leak location and confidence information; when no abnormal leak interval is identified, output no leak status information.

[0080] Furthermore, for the set of candidate leak locations, a refined localization feature vector is constructed, which is represented as follows:

[0081] Where x represents the refined localization feature vector; These represent different refined positioning features. The refined positioning features preferably include the average offset of the abnormal segment at each measuring point, the upstream and downstream pressure difference at the candidate location, the local pressure drop ratio, the dynamic response slope in the initial stage of the anomaly, and the pressure offset sorting code, etc.

[0082] Furthermore, the refined localization features are input into a classification and discrimination model to assist in the discrimination of candidate locations. The auxiliary discrimination model is preferably an ECOC multi-class discrimination model based on support vector machines, but it can also be other supervised learning models suitable for distinguishing multi-class candidate locations. The model outputs the most likely leak location and its confidence level.

[0083] in, To determine the most likely leak location output by the classification model, For classification and discrimination models, in Figure 3 In the specific implementation shown, The values ​​are F1, F2, F3, F4, F5, F6, F7 and F8.

[0084] Furthermore, to improve the stability of the discrimination results, the discrimination results of candidate leak locations can be output separately in multiple consecutive detection windows, and a multi-window voting method can be used to determine the final most likely leak location. When the discrimination results of multiple windows point to the same candidate leak point, the confidence information corresponding to that candidate leak point is increased; when the scores of multiple candidate leak points are close or the discrimination results of multiple windows are scattered, the confidence information is reduced, and the set of candidate leak locations is retained for on-site inspection and confirmation.

[0085] Finally, the output shows the start and end times of the leakage anomaly, the set of candidate leakage locations, the most likely leakage location, and its confidence level. The output can be represented as:

[0086] Where R represents the leak diagnosis and location result. For the set of candidate leak locations, Conf represents the most likely location of the leak, and Conf represents the confidence level.

[0087] Furthermore, through the above S1 to S6, this specific embodiment will Figure 3 The nodes, pipe segments, pressure measuring points, and candidate leak points in the gas drainage pipeline network shown are unified into the same topological mapping relationship. Eight sets of branch pressure difference characteristics are constructed using paired pressure measuring points upstream and downstream of the candidate leak points. These characteristics are then combined with the pressure response characteristics of multiple measuring points P1 to P17 and the topological propagation law to score and assist in the discrimination of candidate leak points. This method can reflect the local pressure difference abrupt change characteristics of leak points and reduce confusion between adjacent leak points or leak points with similar responses by utilizing pipeline network topological constraints, thereby improving the stability and accuracy of leak identification and location results in the gas drainage pipeline network.

[0088] Furthermore, the data preprocessing, stable section identification, anomaly detection, candidate screening, and fine classification steps in this invention can be deployed in a modular manner on host computer software, industrial monitoring platforms, or embedded diagnostic systems to achieve online or offline diagnosis of gas extraction pipeline leakage status.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the method of the present invention, and are not intended to limit it. Although the present invention 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 method described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding method to deviate from the scope defined by the claims of the present invention.

Claims

1. A method for diagnosing and locating leaks in gas extraction pipeline networks based on branch pressure difference characteristics and topological constraints, comprising the following steps: S1: Receives pressure time series data from multiple pressure measuring points in the gas extraction pipeline network, as well as topology mapping information between nodes, pipe sections, pressure measuring points, and candidate leak points; S2: Preprocess the pressure time series data, the preprocessing including one or more of time alignment, missing data compensation, anomaly removal, unit unification and smoothing; S3: Based on the stable operating period, construct an adaptive pressure baseline for each pressure measuring point, calculate the pressure offset of each pressure measuring point relative to the adaptive pressure baseline, and construct a multi-measuring point joint anomaly statistic based on the pressure offset to determine the leakage anomaly range. S4: Within the leakage anomaly range, extract the branch pressure difference characteristics between paired pressure measuring points upstream and downstream of the candidate leakage point and the pressure response characteristics of multiple measuring points to construct leakage pressure characteristics; S5: Based on the topology mapping information, the branch pressure difference characteristics, and the multi-measuring point pressure response characteristics, perform topology constraint scoring on each candidate leak point, and filter the candidate leak location set according to the scoring results; S6: Perform auxiliary discrimination on the candidate leak location set, and output the leak start time, leak end time, candidate leak location set, most likely leak location and confidence information; when no abnormal leak interval is identified, output no leak status information.

2. The method for diagnosing and locating leaks in gas extraction pipeline networks based on branch pressure difference characteristics and topological constraints as described in claim 1, characterized in that, In step S1, the pressure time series data can be real-time data collected on-site, historical data from the industrial monitoring platform, data collected by the host computer, or data imported from external files. The imported external files include CSV, TXT, database export files, or pressure time series data files exported by the monitoring system.

3. The method for diagnosing and locating leaks in gas extraction pipeline networks based on branch pressure difference characteristics and topological constraints as described in claim 1, characterized in that, In step S1, the topology mapping information includes the connection relationship between nodes and pipe segments, the correspondence between pressure measuring points and pipe segments or nodes, the correspondence between candidate leak points and their upstream and downstream pressure measuring points, as well as the adjacency relationship, flow path relationship, confluence node relationship or topological distance relationship between candidate leak points. The gas extraction pipeline network is abstracted into a graph structure composed of a set of nodes and a set of pipe segments, and the degree of association between candidate leak points and each pressure measuring point is determined based on the graph structure.

4. The method for diagnosing and locating leaks in gas extraction pipeline networks based on branch pressure difference characteristics and topological constraints as described in claim 1, characterized in that, In step S2, the preprocessing includes: resampling the pressure data from multiple measurement points according to a unified time axis; interpolating and compensating for short-term missing data; identifying and correcting abrupt anomalies; and unifying pressure data from different sources into the same physical unit. Noise interference is reduced by filtering, moving average, or median smoothing.

5. The method for diagnosing and locating leaks in gas extraction pipeline networks based on branch pressure difference characteristics and topological constraints as described in claim 1, characterized in that... In step S3, the construction of the adaptive pressure baseline includes using a sliding time window to calculate the average pressure, fluctuation degree, and change in adjacent windows of each pressure measuring point in different windows; and determining continuous windows that meet the stability judgment conditions as stable candidate windows. Based on the data within the stable candidate window, a stable operating baseline for each pressure measurement point under the current operating conditions is constructed.

6. The method for diagnosing and locating leaks in gas extraction pipeline networks based on branch pressure difference characteristics and topological constraints as described in claim 1, characterized in that, In step S3, determining the leakage anomaly interval based on the joint anomaly statistics of multiple measuring points includes: calculating the pressure offset of each pressure measuring point relative to the adaptive pressure baseline and its standardized anomaly statistics; determining the proportion of each pressure measuring point exceeding the threshold within the detection window; determining the start of leakage anomaly when the number of joint anomaly measuring points, the proportion of exceeding the threshold, and the duration meet the preset triggering conditions; and determining the end of leakage anomaly when the joint anomaly conditions are continuously not met and the recovery determination time is reached.

7. The method for diagnosing and locating leaks in gas extraction pipeline networks based on branch pressure difference characteristics and topological constraints as described in claim 1, characterized in that, In step S4, the branch pressure difference characteristics are composed of the pressure difference between paired pressure measuring points upstream and downstream of the candidate leak point, the pressure difference change, or the pressure difference offset relative to the stable operating baseline; the multi-measuring point pressure response characteristics include one or more of the following: average offset of abnormal section, peak offset, local pressure drop ratio, dynamic response slope of the initial stage of abnormality, and amplitude ranking information.

8. The method for diagnosing and locating leaks in gas extraction pipeline networks based on branch pressure difference characteristics and topological constraints as described in claim 1, characterized in that, In step S5, the topology constraint scoring includes constructing a scoring function based on the correlation between candidate leak points and adjacent pressure measuring points, the pressure difference change relationship between upstream and downstream paired pressure measuring points of candidate leak points, and the attenuation relationship when the leakage disturbance propagates along the pipeline network. The scoring function is used to rank each candidate leak point and select one or more of the top-scoring locations to form a set of candidate leak locations.

9. The method for diagnosing and locating leaks in gas extraction pipeline networks based on branch pressure difference characteristics and topological constraints as described in claim 1, characterized in that, In step S6, the auxiliary discrimination is achieved by combining the topological constraint scoring results with the ECOC multi-class discrimination model based on support vector machine; wherein, the topological constraint scoring results are used to form a set of candidate leak locations, and the ECOC multi-class discrimination model based on support vector machine is used to further distinguish the locations in the candidate leak location set that are topologically adjacent, close in distance, or have similar pressure responses, and output the most likely leak location.

10. The method for diagnosing and locating leaks in gas extraction pipeline networks based on branch pressure difference characteristics and topological constraints as described in claim 1, characterized in that, In step S6, the confidence information is determined by one or more of the following: topological constraint score, auxiliary discrimination model output score, score difference between candidate leakage locations, and multi-window voting results; the output results are used for gas extraction pipeline network leakage alarm, inspection prompt, maintenance location, valve adjustment, or negative pressure optimization allocation.

Images