Gas pipe network leakage early warning method and system based on side cloud collaboration

Abstract

The invention relates to the field of gas pipe network leakage early warning, in particular to a gas pipe network leakage early warning method based on edge cloud collaboration, and the method comprises the steps: responding to a leakage monitoring instruction, obtaining pipe network operation data, generating monitoring configuration parameters based on the pipe network operation data, and transmitting the monitoring configuration parameters to an edge gateway; receiving a historical training curve library, performing feature extraction on each historical training curve in the historical training curve library to obtain a training feature value, training the initial verification model based on the training feature value, and generating an early warning verification model; receiving pipe network early warning data obtained by the edge gateway based on the monitoring configuration parameters, sending the pipe network early warning data to an early warning verification model for verification, and obtaining early warning time series data, the early warning time series data including preorder data and post-order data; and obtaining an early warning verification result based on the preorder data and the postorder data, generating a leakage early warning interface when the early warning verification result is leakage, and sending the leakage early warning interface to a maintenance end.

Description

Technical Field

[0001] The present invention relates to the field of gas pipeline network leakage warning, and in particular to a gas pipeline network leakage warning method and system based on edge-cloud collaboration. Background Art

[0002] The gas pipeline network is the core of urban energy supply, and its safety is directly related to public safety and urban operation. Once a leak occurs, the flammable and explosive gas will form an explosive mixture with the air, bringing serious fire and poisoning risks. In addition, gas leaks will also lead to energy waste and increased operating costs, and the pollution that is difficult to recover can easily cause long-term impacts on the environment and health. Therefore, real-time monitoring of the gas pipeline network and timely handling of leakage incidents are important measures to ensure urban safety, maintain the environment and reduce economic losses.

[0003] Although the current gas pipeline network monitoring method relies on the rapid development of information technology and the Internet of Things, and has achieved a transition from traditional manual inspections to online monitoring, for small-scale or slow leaks, the data change is not large, and insufficient monitoring sensitivity may lead to missed reports. In addition, the complex pipeline structure and external environment may also interfere with the monitoring data. Under high-sensitivity conditions, it is easy to trigger false alarms due to slight fluctuations. False alarms or missed reports will pose a potential threat to the safe operation of the gas pipeline network. Summary of the invention

[0004] The present invention provides a gas pipeline network leakage warning method and system based on edge-cloud collaboration, which can reduce false alarms or missed alarms caused by tiny data fluctuations and interference from external factors, and improve the efficiency, reliability and accuracy of pipeline network leakage warning.

[0005] In order to solve the above technical problems, a technical solution adopted by the present invention is: to provide a gas pipeline network leakage early warning method based on edge-cloud collaboration, the method comprising: responding to a leakage monitoring instruction, obtaining pipeline network operation data, generating monitoring configuration parameters based on the pipeline network operation data, and sending the monitoring configuration parameters to an edge gateway; receiving a historical training curve library, performing feature extraction on each historical training curve in the historical training curve library, obtaining training feature values, training an initial verification model based on the training feature values, and generating an early warning verification model; Receive the pipe network warning data obtained by the edge gateway based on the monitoring configuration parameters, send the pipe network warning data to the warning verification model for verification, and obtain warning time series data, wherein the warning time series data includes preceding data and succeeding data; obtain a warning verification result based on the preceding data and the succeeding data, and when the warning verification result is a leak, generate a leakage warning interface, and send the leakage warning interface to the maintenance end.

[0006] Optionally, responding to the leakage monitoring instruction, acquiring the pipe network operation data, generating monitoring configuration parameters based on the pipe network operation data, and sending the monitoring configuration parameters to the edge gateway include: In response to the leakage monitoring instruction, the pipe network operation data sent by the edge gateway is received, and the pipe network operation data includes point data, pipeline data and environmental interference data; the preset parameter configuration strategy is called to identify the pipeline data, obtain transportation attribute data and historical monitoring data, and determine the basic configuration parameters according to the historical monitoring data; based on the point data, regional terrain data, land use data and energy supply depth data are obtained, and the preset parameter impact assessment table is called to respectively obtain the point assessment association values ​​between the transportation attribute data and the regional terrain data, the land use data and the energy supply depth data, and the point correction value is determined based on each point assessment association value; based on the environmental interference data, meteorological condition data and traffic flow data are obtained, and the environmental assessment association values ​​between the transportation attribute data and the meteorological condition data and the traffic flow data are respectively obtained according to the preset parameter impact assessment table, and the environmental correction value is determined based on each environmental assessment association value; based on the point correction value, the environmental correction value and the basic configuration parameters, the monitoring configuration parameters are generated, and the monitoring configuration parameters are sent to the edge gateway.

[0007] Optionally, a historical training curve library is received, feature extraction is performed on each historical training curve in the historical training curve library, training feature values ​​are obtained, an initial verification model is trained based on the training feature values, and an early warning verification model is generated, including: responding to a model training instruction, receiving a historical training curve library, identifying the historical training database, and obtaining a state classification label for each historical training curve; determining a labeling state of the corresponding historical training curve based on the state classification label, wherein the labeling state includes a leakage state and a non-leakage state; feature extraction is performed on the historical training curves of the leakage state and the non-leakage state, respectively, to obtain corresponding training feature values, wherein the training feature values ​​include similarity feature values, linear fitting feature values, fluctuation feature values, and discrete feature values; a feature training matrix is ​​generated based on the similarity feature values, the linear fitting feature values, the fluctuation feature values, and the discrete feature values; a state label matrix is ​​constructed based on the feature training matrix and each state classification label, wherein the state classification label of the state label matrix corresponds to the order of the historical training curves in the feature training matrix; an initial verification model is trained based on the feature training matrix and the state label matrix, and an early warning verification model is generated.

[0008] Optionally, a feature training matrix is ​​generated based on the similar eigenvalues, the linear fitting eigenvalues, the fluctuating eigenvalues ​​and the discrete eigenvalues, including: constructing a leakage feature matrix based on the similar eigenvalues, the linear fitting eigenvalues, the fluctuating eigenvalues ​​and the discrete eigenvalues ​​corresponding to the historical training curve of the leakage state; constructing a non-leakage feature matrix based on the similar eigenvalues, the linear fitting eigenvalues, the fluctuating eigenvalues ​​and the discrete eigenvalues ​​corresponding to the historical training curve of the non-leakage state; and stacking the leakage feature matrix and the non-leakage feature matrix to generate a feature training matrix.

[0009] Optionally, the historical training curve of the leakage state is a leakage pressure curve, and feature extraction is performed on the historical training curve of the leakage state to obtain training feature values, including: obtaining each leakage pressure data point based on the leakage pressure curve, combining any two leakage pressure data points to generate a pressure difference similarity analysis group; calling a preset similarity analysis function to obtain the pressure difference similarity value of each pressure difference similarity analysis group, and generating a pressure difference similarity matrix based on each pressure difference similarity value; obtaining the maximum pressure difference interval based on the pressure difference similarity matrix, and determining the similarity feature value according to the maximum pressure difference interval.

[0010] Optionally, feature extraction is performed on each historical training curve in the historical training curve library to obtain training feature values, and the method also includes: identifying the leakage pressure curve to obtain a warning time label corresponding to each leakage pressure data point; calling a preset linear regression model to fit each leakage pressure data point and the corresponding warning time label to obtain a tilt coefficient and an initial offset, and determining the tilt coefficient as a tilt feature value; determining the initial offset as an intercept feature value, and obtaining a linear fitting feature value based on the tilt feature value and the intercept feature value.

[0011] Optionally, feature extraction is performed on each historical training curve in the historical training curve library to obtain training feature values, and the method also includes: performing mean calculation on the leakage pressure value corresponding to each leakage pressure data point to obtain the pressure average value, and locating the fluctuation baseline based on the pressure average value; determining the positive deviation area and the negative deviation area based on the fluctuation baseline, and calling a preset area solution strategy to obtain the positive deviation area of ​​the positive deviation area and the negative deviation area of ​​the negative deviation area; obtaining an area difference based on the positive deviation area and the negative deviation area, and determining the area difference as the fluctuation feature value.

[0012] Optionally, feature extraction is performed on each historical training curve in the historical training curve library to obtain training feature values, and the method also includes: performing difference calculations between the pressure average value and the leakage pressure value corresponding to each leakage pressure data point to obtain the pressure difference fluctuation; calculating the square of each pressure difference fluctuation to obtain a single-point fluctuation variance, obtaining a cumulative fluctuation variance based on the sum of each single-point fluctuation variance, and determining the cumulative fluctuation variance as a discrete feature value.

[0013] Optionally, obtaining a warning verification result based on the preceding data and the succeeding data includes: performing feature extraction based on the preceding data and the succeeding data to obtain preceding feature values ​​and succeeding feature values, respectively; integrating the preceding feature values ​​and the succeeding feature values ​​to obtain a warning correlation value, comparing the warning correlation value with the training feature value to obtain a warning verification result.

[0014] To solve the above technical problems, another technical solution adopted by the present invention is: to provide a gas pipeline leakage warning system based on edge-cloud collaboration, the system comprising: a parameter configuration module, used to respond to leakage monitoring instructions, obtain pipeline operation data, generate monitoring configuration parameters based on the pipeline operation data, and send the monitoring configuration parameters to the edge gateway; a model acquisition module, used to receive a historical training curve library, perform feature extraction on each historical warning curve in the historical training curve library, obtain training feature values, train the initial verification model based on the training feature values, and generate a warning verification model; a timing verification module, used to receive the pipeline warning data obtained by the edge gateway based on the monitoring configuration parameters, send the pipeline warning data to the warning verification model for verification, and obtain warning time series data, the warning time series data including preceding data and succeeding data; an interface generation module, used to obtain a warning verification result based on the preceding data and the succeeding data, and when the warning verification result is a leak, generate a leakage warning interface, and send the leakage warning interface to the maintenance end.

[0015] Different from the prior art, the present invention provides a gas pipeline leakage warning method based on edge-cloud collaboration, including: responding to leakage monitoring instructions, obtaining pipeline operation data, generating monitoring configuration parameters based on the pipeline operation data, and sending the monitoring configuration parameters to the edge gateway; receiving a historical training curve library, performing feature extraction on each historical training curve in the historical training curve library, obtaining training feature values, training an initial verification model based on the training feature values, and generating a warning verification model; receiving pipeline warning data obtained by the edge gateway based on the monitoring configuration parameters, sending the pipeline warning data to the warning verification model for verification, and obtaining warning time series data, the warning time series data including preceding data and succeeding data; obtaining a warning verification result based on the preceding data and the succeeding data, and when the warning verification result is a leak, generating a leakage warning interface, and sending the leakage warning interface to the maintenance end. The present invention realizes efficient early warning of gas pipeline leakage by combining edge computing with cloud computing. After the edge device performs real-time preliminary processing and analysis on the pipeline data, the pipeline warning data with reduced noise and redundant information is uploaded to the cloud. The cloud uses powerful computing and storage capabilities to further analyze the pipeline warning data in depth to obtain early warning verification results. When the early warning verification result is a leak, a leakage early warning interface is generated and sent to the maintenance end, thereby improving the accuracy of pipeline leakage early warning. The cloud adopts an early warning verification model in the analysis process, which can accurately process time series data and discover minor anomalies, reducing underreporting caused by small-scale or slow leaks and false alarms caused by external interference, thereby improving the efficiency and reliability of pipeline leakage early warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] One or more embodiments are exemplarily described by corresponding drawings, which do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and the figures in the drawings do not constitute proportional limitations unless otherwise stated.

[0017] Figure 1 It is a flow chart of a gas pipeline network leakage early warning method based on edge-cloud collaboration provided by an embodiment of the present invention; Figure 2 It is a schematic diagram of an edge-cloud collaborative architecture provided by an embodiment of the present invention; Figure 3 It is a schematic diagram of a pressure curve corresponding to a gas pipeline network provided by an embodiment of the present invention; Figure 4 It is a schematic diagram of a training process of an early warning verification model provided by an embodiment of the present invention; Figure 5 It is a schematic diagram of obtaining training feature values ​​based on a historical training curve provided by an embodiment of the present invention; Figure 6 It is a structural diagram of a gas pipeline leakage warning system based on edge-cloud collaboration provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0019] It should be noted that, if there is no conflict, the various features in the embodiments of the present invention can be combined with each other, and all are within the protection scope of the present invention. In addition, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flow chart, in some cases, the steps shown or described can be performed in a different order from the module division in the device schematic diagram or the order in the flow chart.

[0020] The terms "first", "second", "third", "fourth", etc. (if any) in the description and claims of the present invention and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in sequences other than those illustrated or described herein.

[0021] It should be understood that in the present invention, "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or apparatuses.

[0022] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0023] See also Figure 1 , Figure 1 It is a flow chart of a gas pipeline network leakage warning method based on edge-cloud collaboration provided by an embodiment of the present invention, including steps S1 to S4, which are specifically as follows: S1. Respond to a leakage monitoring instruction, obtain pipe network operation data, generate monitoring configuration parameters based on the pipe network operation data, and send the monitoring configuration parameters to an edge gateway.

[0024] Among them, edge-cloud collaboration is a technical architecture that combines edge computing and cloud computing to jointly realize leakage warning of gas pipeline networks; the gas pipeline network refers to the pipeline system for transporting natural gas, including pipelines, valves, monitoring sensors, etc.; the leakage monitoring instruction refers to the instruction issued by the user or the system, which aims to start the leakage monitoring process to detect leakage events in the gas pipeline network in real time; the pipeline network operation data represents the operating status of the gas pipeline network, reflecting the working conditions and status of the pipeline network; the monitoring configuration parameters represent the operating standards and conditions of the monitoring system set by the cloud for the edge gateway, such as monitoring frequency, warning threshold, etc., which affect the system's monitoring capabilities and response speed; the edge gateway represents a device for edge computing, which can receive data collected by sensors and preliminarily determine whether a leak has occurred.

[0025] See also Figure 2 , Figure 2 Schematic diagram of an edge-cloud collaborative architecture provided by an embodiment of the present invention. The edge-cloud collaborative architecture includes a cloud and an edge gateway. Figure 2 Three edge gateways are shown, which are usually located at the edge of the network, close to the source of data generation. They are mainly responsible for real-time data processing and low-latency tasks, such as rapid data judgment and early warning. They usually process local data, can perform edge computing and respond quickly to local events. The edge gateway is also responsible for uploading the initially processed data to the cloud for further analysis and storage. Each edge gateway ID represents the unique identifier of each edge gateway.

[0026] The cloud consists of data servers and computing servers. The data servers are responsible for data storage and recording, while the computing servers process and analyze data. The computing servers can further optimize configuration parameters based on the data uploaded by the edge gateway to increase the sensitivity of the early warning. They are configured with pre-trained machine learning models, which can conduct in-depth learning and analysis of the received data to form a more accurate data judgment model and filter the early warning information uploaded by the edge gateway to reduce false alarms.

[0027] By processing most real-time and low-latency tasks at the edge, and leveraging the powerful computing and storage capabilities of the cloud for in-depth analysis and parameter optimization, the edge-cloud collaborative architecture can optimize resource allocation across the entire network and improve the system's monitoring accuracy and response sensitivity to leakage incidents.

[0028] It should be noted that the operating environment and status of different gas pipeline networks are different. The cloud adjusts the monitoring parameter settings of the edge gateway according to the specific differences to ensure that each edge gateway can meet the monitoring requirements in the corresponding scenarios.

[0029] In some embodiments, step S1 (responding to the leakage monitoring instruction, obtaining the pipe network operation data, generating the monitoring configuration parameters based on the pipe network operation data, and sending the monitoring configuration parameters to the edge gateway) specifically includes steps S11-S15: S11. Respond to the leakage monitoring instruction and receive the pipe network operation data sent by the edge gateway, wherein the pipe network operation data includes point data, pipeline data and environmental interference data.

[0030] Among them, the point data represents the specific location of the gas pipeline network, which can be used to locate the monitoring area and related geographical attribute information; the pipeline data represents the physical and operating characteristics of the pipeline; the environmental interference data represents the data in the external environment around the pipeline that may affect the operation of the gas pipeline network.

[0031] Specifically, after receiving the instruction from the user or system to start the leak monitoring process, the cloud will work with the edge gateway through wireless communication to receive and process the relevant data collected by the edge gateway on the operating status of the gas pipeline network. It should be noted that the acquired point data, pipeline data and environmental interference data can be used as input data for subsequent parameter configuration and can provide a basis for parameter configuration.

[0032] S12, calling a preset parameter configuration strategy to identify the pipeline data, obtain transportation attribute data and historical monitoring data, and determine basic configuration parameters according to the historical monitoring data.

[0033] Among them, the preset parameter configuration strategy represents a pre-set parameter configuration scheme; the transportation attribute data represents the basic characteristics and design parameters of the gas pipeline network, such as the pipeline material; the historical monitoring data represents past monitoring records and data, including historical monitoring parameters during pipeline operation; the basic configuration parameters represent the preliminary monitoring parameters set by the cloud for the edge gateway, that is, the parameters set based on the average value of historical monitoring parameters.

[0034] For example, pipeline data is identified and the material of the pipeline is obtained as steel pipe, which is the transportation attribute data. The average gas pressure calculated based on past monitoring records and data is 4.5 bar, and the average gas flow rate is 2.8 m / s. The pressure warning threshold can be set to 5.5 bar and the flow rate warning threshold can be set to 3.5 m / s based on the average gas pressure. The basic configuration parameters include pressure warning threshold, flow rate warning threshold and other parameters.

[0035] It should be noted that the basic configuration parameters generated by the system can be used as a benchmark for subsequent adjustment of parameters based on environmental and geographical conditions.

[0036] S13. Based on the point data, regional terrain data, land use data and energy supply depth data are obtained, and the preset parameter impact assessment table is called to respectively obtain the point assessment correlation values ​​between the transportation attribute data and the regional terrain data, the land use data and the energy supply depth data, and the point correction value is determined based on each point assessment correlation value.

[0037] Among them, the regional terrain data represents the terrain characteristics of the area where the gas pipeline network is located, such as plains, mountains, river valleys, etc.; the land use data represents the land use conditions in the area where the gas pipeline network is located, such as cities, farmlands and industrial areas, etc.; the energy supply depth data represents the underground burial depth of the gas pipeline network, such as: deep, moderate, shallow; the preset parameter impact assessment table represents a pre-set table for quantifying the impact of different regional terrain, land use and energy supply depth on transportation properties; the point evaluation association value is the impact value between regional terrain, land use, energy supply depth and transportation properties, and the monitoring configuration is adjusted through these impact values; the point correction value, that is, the sum of the point evaluation association values, represents the parameter adjustment value set to fit the specific conditions of the actual point, to ensure that the monitoring system matches the actual geographical conditions.

[0038] For example, the regional terrain data corresponding to the gas network is plain, the land use data is urban, and the energy supply depth data is shallow. The system finds through the preset parameter impact assessment table that for the pressure warning threshold, the point assessment correlation value between the plain (regional terrain) and the steel pipe (transportation attribute) is 0.3 bar, the point assessment correlation value between the city (land use) and the steel pipe (transportation attribute) is 0.5 bar, and the point assessment correlation value between the shallow (energy supply depth) and the steel pipe (transportation attribute) is -0.3 bar, so the point correction value is 0.5 bar. The point correction value corresponding to the velocity warning threshold can be obtained using the same idea, such as the point correction value corresponding to the velocity warning threshold is 0.2 m / s.

[0039] It should be noted that the geographical environment of each gas pipeline network is different and may have different monitoring requirements. By obtaining the point correction value, the basic configuration parameters can be refined and adjusted to match the actual needs of each gas pipeline network.

[0040] S14. Obtain meteorological condition data and traffic flow data based on the environmental interference data, obtain environmental assessment association values ​​between the transportation attribute data and the meteorological condition data and the traffic flow data respectively according to the preset parameter impact assessment table, and determine environmental correction values ​​based on each environmental assessment association value.

[0041] Among them, the meteorological condition data represents the weather information of the area where the gas pipeline network is located, such as temperature, rainfall and wind speed, etc., and describes the climate change in the area around the gas pipeline; the traffic flow data represents the traffic data near the gas pipeline network, such as busy traffic, smooth traffic, etc.; the environmental assessment associated value is a numerical value used to quantify the impact of meteorological conditions and traffic flow on the transportation properties of the gas pipeline network; the environmental correction value is the sum of each environmental assessment associated value, which represents the parameter adjustment value set to compensate for the impact of the external environment on the monitoring system.

[0042] For example, the meteorological condition data corresponding to the gas pipeline network is sunny, and the traffic flow data is busy traffic. The system finds the pressure warning threshold through the preset parameter impact assessment table. The environmental assessment correlation value between sunny day (meteorological condition) and steel pipe (transportation attribute) is 0.4 bar, and the environmental assessment correlation value between busy traffic (traffic flow) and steel pipe (transportation attribute) is 0.2 bar, so the environmental correction value is 0.6 bar. The environmental correction value corresponding to the velocity warning threshold can be obtained using the same idea, such as the environmental correction value corresponding to the velocity warning threshold is 0.3 m / s.

[0043] It should be noted that fluctuations in the external environment, such as weather changes and traffic conditions, may affect the operation of the gas pipeline network. Through the environmental correction value, the sensitivity and response speed of the monitoring system can be dynamically adjusted to reduce the impact of external interference.

[0044] S15. Generate monitoring configuration parameters based on the point correction value, the environment correction value and the basic configuration parameters, and send the monitoring configuration parameters to an edge gateway.

[0045] The monitoring configuration parameters represent a set of parameters generated based on the point correction value, the environment correction value and the basic configuration parameters, and are used to guide the monitoring behavior and operation standard of the edge gateway.

[0046] For example, after configuration adjustment, the pressure warning threshold is 5.5 bar + 0.5 bar + 0.6 bar, that is, 6.6 bar, and the flow rate warning threshold is 3.5 m / s + 0.2 m / s + 0.3 m / s, that is, 4 m / s. In the generated monitoring configuration parameters, the pressure warning threshold is 6.6 bar and the flow rate warning threshold is 4 m / s.

[0047] It should be noted that the monitoring configuration parameters finally generated can ensure that the monitoring behavior of the edge gateway can accurately match the actual environment and operating status to achieve the best monitoring effect.

[0048] S2. Receive a historical training curve library, perform feature extraction on each historical training curve in the historical training curve library, obtain training feature values, train an initial verification model based on the training feature values, and generate an early warning verification model.

[0049] Among them, the historical training curve library represents a database containing multiple past gas pipeline network monitoring data curves, which is the basis of model training; the historical training curve represents a single monitoring data curve extracted from the historical training curve library, representing the monitoring information of the pipeline network operation within a certain period of time; the training feature value represents the feature value extracted from the historical training curve, representing various important information of the curve; the initial verification model represents a basic model that has not been trained or has only been trained a small amount, which can be a bidirectional long short-term memory recurrent neural network (Bidirectional LongShort-Term Memory, BiLSTM for short) model; the early warning verification model is based on the initial verification model, using the training feature values ​​extracted from historical data for further training and optimization, so as to generate a model for verifying whether the early warning data uploaded to the cloud is accurate.

[0050] It should be noted that in the warning data uploaded by the edge gateway, there may be false alarms or missed alarms due to small-scale or slow leakage, sensor noise, environmental factors, etc. By training to obtain the warning verification model, it is convenient to further verify the warning data uploaded by the edge gateway to confirm whether there is a real problem, screen out false alarms or missed alarms, and ensure the accuracy of the warning data.

[0051] It is worth noting that the monitoring data in different monitoring data curves in the gas pipeline network usually show specific rules. Understanding these rules is crucial for the establishment of machine learning models. The model needs to verify the early warning data uploaded to the cloud based on the above rules.

[0052] See also Figure 3 , Figure 3 It is a schematic diagram of a pressure curve corresponding to a gas pipeline network provided in an embodiment of the present invention. Figure 3 -a is the pressure data curve under normal operating conditions. The pressure value fluctuates within a certain range, showing the regular changes in daily life. Figure 3 -b The curve in the leftmost box shows the change in pressure value when the pipeline leaks, that is, the pressure drops suddenly when the leak occurs. Figure 3 -a pressure value fluctuations are slow and periodic, while Figure 3 The pressure value change caused by the leakage in -b is rapid and drastic. This sudden change requires the model to focus on identifying it.

[0053] See also Figure 4 , Figure 4 It is a schematic diagram of a training process of an early warning verification model provided by an embodiment of the present invention.

[0054] In some embodiments, step S2 (receiving a historical training curve library, extracting features from each historical training curve in the historical training curve library, obtaining training feature values, training an initial verification model based on the training feature values, and generating an early warning verification model) specifically includes steps S21-S26: S21. Respond to the model training instruction, receive the historical training curve library, identify the historical training database, and obtain the state classification label of each historical training curve.

[0055] Among them, the model training instruction represents a command triggered by the system, which can instruct the cloud to start training the model based on the historical training curve library; the state classification label represents the state category to which the curve belongs.

[0056] It should be noted that in order to train the model, it is necessary to use manual labeling and other methods to mark and classify each historical training curve in the historical training curve library in advance, and clearly set the corresponding state classification label for each historical training curve, so that the model can learn the characteristics under different states based on these labels, thereby accurately verifying the early warning data.

[0057] S22: Determine a labeling state of a corresponding historical training curve based on the state classification label, where the labeling state includes a leakage state and a non-leakage state.

[0058] Among them, the marked state represents the pre-defined operating state of the gas pipeline network, which is used to indicate the specific category to which the historical training curve belongs; the leakage state represents the state of the gas pipeline network when leakage occurs during operation; the non-leakage state represents the state of the gas pipeline network when it is operating normally or without leakage.

[0059] S23, performing feature extraction on the historical training curves of the leakage state and the non-leakage state respectively to obtain corresponding training feature values, wherein the training feature values ​​include similarity feature values, linear fitting feature values, fluctuation feature values ​​and discrete feature values.

[0060] Among them, the similarity eigenvalue represents the eigenvalue used to reflect the maximum similarity difference between data points; the linear fitting eigenvalue represents the eigenvalue that describes the overall linear change trend of the curve; the fluctuation eigenvalue represents the eigenvalue used to quantify the degree of fluctuation of the curve, which can reveal the degree of imbalance of the curve in positive and negative fluctuations; the discrete eigenvalue represents the eigenvalue used to measure the fluctuation amplitude of the curve over time.

[0061] It should be noted that feature extraction is to simplify complex historical data into numerical values ​​that are easy for the model to process. The training feature values ​​obtained through feature extraction can describe the shape and trend of the curve and help the model identify different operating conditions of the gas pipeline network.

[0062] Figure 5 It is a schematic diagram of obtaining training feature values ​​based on a historical training curve provided by an embodiment of the present invention.

[0063] In some embodiments, the historical training curve of the leakage state is a leakage pressure curve, and feature extraction is performed on the historical training curve of the leakage state to obtain training feature values, specifically including steps A1-A3: A1. Acquire each leakage pressure data point based on the leakage pressure curve, combine any two leakage pressure data points, and generate a pressure difference similarity analysis group.

[0064] Among them, the leakage pressure curve represents a curve of the internal pressure of the pipeline network changing with time when a leak occurs in the gas pipeline network; the leakage pressure data points represent the discrete pressure values ​​in the leakage pressure curve, and each data point represents the pressure value at a certain moment; the pressure difference similarity analysis group represents a set composed of any two leakage pressure data points, that is, each pressure difference similarity analysis group contains two different leakage pressure data points, and the number of the pressure difference similarity analysis groups can be adjusted according to the number of leakage pressure data points to obtain multiple pressure difference similarity analysis groups.

[0065] It should be noted that a single data point is not sufficient to reveal the changing pattern of pressure values. Obtaining a pressure difference similarity analysis group can compare the similarity of pressure data points at different times, capture the regularity and abnormal characteristics of pressure value changes, and improve the accuracy of model verification.

[0066] A2. Calling a preset similarity analysis function to obtain the pressure difference similarity value of each pressure difference similarity analysis group, and generating a pressure difference similarity matrix based on each pressure difference similarity value.

[0067] Among them, the preset similarity analysis function is a function for calculating the similarity between different data points, which can be specifically a Gaussian kernel function; the pressure difference similarity value represents the degree of proximity between two leakage pressure data points in terms of characteristics; the pressure difference similarity matrix represents the similarity value between every two leakage pressure data points, which is a two-dimensional matrix.

[0068] See also Figure 3 -b, the leaked data in the leftmost box are combined into the leaked pressure data PL, and the non-leaked data in the three right boxes are combined into the non-leaked pressure data PN, so: , P is an n×m data matrix, and T represents the transpose of the matrix, that is, the rows and columns are interchanged.

[0069] Specifically, the pressure difference similarity value is calculated by the following formula: ;in, Indicates the leak pressure data point Data points with leak pressure The similarity between them, that is, the pressure difference similarity value, i, j is the corresponding i row and j row in P represents the Euclidean distance. is the Gaussian kernel bandwidth, and e is the base of the natural logarithm which is also a constant.

[0070] It should be noted that the higher the similarity value, the closer the pressure difference change between the two leakage pressure data points is, that is, the higher the degree of closeness in the characteristics, and the lower the similarity value, the farther the pressure difference change between the two leakage pressure data points is, that is, the lower the degree of closeness in the characteristics. By obtaining the pressure difference similarity value, it can be known at which moments the pressure change patterns are similar, thereby identifying the characteristics of the gas pipeline network leakage. The application of Gaussian kernel function is well known to those skilled in the art, and the present invention will not be repeated here.

[0071] The resulting similarity matrix is ​​as follows: Each element in the matrix represents the pressure difference similarity value between two leakage pressure data points.

[0072] A3. Obtaining a maximum pressure difference interval based on the pressure difference similarity matrix, and determining the similarity eigenvalue according to the maximum pressure difference interval.

[0073] Among them, the maximum pressure difference interval represents the maximum interval between two similar values ​​in the pressure difference similarity matrix. The interval represents the difference between the two most dissimilar data points. Obtaining this interval can determine the key time point in the leakage process, that is, the moment when the pressure value changes most significantly before and after the leakage, which may include the most significant pressure value change characteristics in the leakage process, such as key nodes such as the start, aggravation, or end of the leakage; the similarity characteristic value reflects the key moment of the pressure value change in the leakage event, which can be used as an important basis for the model to identify gas pipeline network leaks.

[0074] Specifically, the leakage characteristic is expressed as follows (for the i-th leakage), which is obtained when the following formula takes the maximum value: ;

[0075] Take similar eigenvalues ​​here: It should be noted that see Figure 5 -a, obtain similar eigenvalues ​​to quantify and analyze the similarity between leakage pressure data points, and use the similar eigenvalues ​​to train the model. The model can identify key change points in the monitoring curve and warn of future leakage trends in the monitoring curve, thereby improving the accuracy of the warning data.

[0076] In some embodiments, feature extraction is performed on each historical training curve in the historical training curve library to obtain training feature values, which specifically includes steps B1-B3: B1. Identify the leakage pressure curve and obtain the warning time label corresponding to each leakage pressure data point.

[0077] The warning time label indicates the time information corresponding to each leakage pressure value marked on the leakage pressure data point, and usually indicates the timestamp corresponding to each leakage pressure value during the leakage process.

[0078] It should be noted that the warning time label is used to track the time when the pressure value changes, so as to determine the change of the pressure value over time during the leakage process. The warning time label can provide data support for the subsequent linear regression analysis.

[0079] B2. Retrieve a preset linear regression model to fit each leakage pressure data point and the corresponding warning time label, obtain a tilt coefficient and an initial offset, and determine the tilt coefficient as a tilt characteristic value.

[0080] Among them, the preset linear regression model is a common statistical model, such as the least squares method, which can be used to analyze the linear relationship between each leakage pressure data point and its corresponding warning time label to obtain the changing trend of the leakage pressure value over time; the inclination coefficient represents the slope of the fitting straight line in the linear regression; the initial offset represents the intercept of the fitting straight line on the time axis, representing the pressure value at time 0; the inclination eigenvalue represents the inclination coefficient (slope) obtained by fitting, and the eigenvalue is used to represent the rate of change of the pressure value during the leakage process.

[0081] Specifically, fitting means finding a straight line that fits all leakage pressure data points as much as possible through a preset linear regression model, and then describing the relationship between leakage pressure values ​​and time according to a mathematical formula to obtain the slope and intercept of the fitting line, see Figure 5 -b,.

[0082] It should be noted that the larger the inclination coefficient (inclination characteristic value), the faster the pressure drops or rises, indicating that the pressure changes more drastically during leakage.

[0083] B3. Determine the initial offset as the intercept eigenvalue, and obtain a linear fitting eigenvalue based on the tilt eigenvalue and the intercept eigenvalue.

[0084] The tilt characteristic value represents the initial offset (intercept) obtained by fitting, which is used to reflect the starting pressure of the leakage pressure curve and can provide information about the initial state of the gas network when the leakage starts; the linear fitting characteristic value represents the slope and intercept of the fitting line.

[0085] Specifically, each leakage pressure data point and the corresponding warning time label are fitted, and the slope and intercept are obtained by the following formula: P = at + b; The slope a is taken as the inclined eigenvalue and the intercept b is taken as the intercept eigenvalue, that is: fi2=a; Fi3=b; It should be noted that by extracting the tilt eigenvalue and the intercept eigenvalue and using these two features as the input of the model, the model can be trained to better understand the pressure change trend of the curve, so that the model can generate more accurate warning data in actual application.

[0086] In some embodiments, feature extraction is performed on each historical training curve in the historical training curve library to obtain training feature values, which specifically also includes steps C1-C3: C1. Calculate the mean of the leakage pressure values ​​corresponding to each leakage pressure data point to obtain a pressure average value, and locate the fluctuation baseline based on the pressure average value.

[0087] Among them, the leakage pressure value represents the pressure value at each moment in the leakage pressure curve; the pressure average value represents the overall level of the pressure value during the leakage process; and the fluctuation baseline represents the dividing line used to divide different fluctuation conditions in the leakage pressure curve.

[0088] Specifically, positioning means determining the horizontal line corresponding to the pressure average value as the fluctuation reference line. It should be noted that locating the fluctuation baseline can separate the fluctuation behavior on the pressure curve into upper and lower parts, providing a reference for subsequent identification of which pressure values ​​are above or below the baseline.

[0089] C2. Determine a positive deviation area and a negative deviation area based on the fluctuation baseline, and call a preset area solution strategy to obtain the positive deviation area of ​​the positive deviation area and the negative deviation area of ​​the negative deviation area.

[0090] Among them, the positive deviation area represents all parts of the leakage pressure curve that are higher than the fluctuation baseline; the negative deviation area represents all parts of the leakage pressure curve that are lower than the fluctuation baseline; the preset area solution strategy represents an algorithm or method for calculating the area of ​​a region, such as calculation by integration or geometric method; the positive deviation area represents the area value calculated based on the positive deviation area, which is used to reflect the degree of positive fluctuation of the pressure value relative to the baseline during the leakage process; the negative deviation area represents the area value calculated based on the negative deviation area, which is used to reflect the degree of negative fluctuation of the pressure value relative to the baseline during the leakage process.

[0091] It should be noted that by calculating the positive deviation area and the negative deviation area, the fluctuation amplitude of the pressure value in different directions can be quantified. The larger the positive deviation area, the greater the fluctuation amplitude of the pressure above the baseline; the larger the negative deviation area, the greater the fluctuation amplitude of the pressure below the baseline.

[0092] C3. Obtain an area difference based on the positive bias area and the negative bias area, and determine the area difference as a fluctuation characteristic value.

[0093] The area difference is the difference between the positive bias area and the negative bias area, reflecting the overall asymmetry of the pressure fluctuation. Figure 5 -c.

[0094] It should be noted that the fluctuation characteristic value is calculated through the area difference, and the model can be trained to understand the fluctuation characteristics of the pressure value during the leakage process.

[0095] In some embodiments, feature extraction is performed on each historical training curve in the historical training curve library to obtain training feature values, which specifically also includes steps D1-D2: D1. Calculate the difference between the pressure average value and the leakage pressure value corresponding to each leakage pressure data point to obtain the pressure difference fluctuation amount.

[0096] Among them, the difference calculation means subtracting the pressure average value from the pressure value of each leakage pressure data point; the pressure difference fluctuation amount represents the degree of deviation of the leakage pressure value relative to the average value, and the pressure difference fluctuation amount can be positive or negative. A positive value indicates that the leakage pressure value is higher than the average value, and a negative value indicates that the leakage pressure value is lower than the average value.

[0097] D2. Calculate the square of each pressure difference fluctuation to obtain the single-point fluctuation variance, obtain the cumulative fluctuation variance based on the sum of the single-point fluctuation variances, and determine the cumulative fluctuation variance as a discrete eigenvalue.

[0098] Among them, each single-point fluctuation variance measures the degree of fluctuation of the data point relative to the pressure average value. The larger the single-point fluctuation variance, the higher the degree to which the pressure value of the point deviates from the average value, and the more violent the fluctuation. The cumulative fluctuation variance is the result obtained by summing up all the single-point fluctuation variances, which represents the overall fluctuation degree of the entire leakage pressure curve relative to the pressure average value. The larger the cumulative fluctuation variance, the more violent the fluctuation of the curve. The cumulative fluctuation variance reflects the fluctuation of the curve at different time points.

[0099] It is worth noting that the squaring process eliminates the influence of positive and negative signs, ensuring that all single-point fluctuation variances are positive so that they can be used for subsequent cumulative calculations. The squared result amplifies larger deviations, reflecting more significant fluctuation characteristics.

[0100] The discrete eigenvalues ​​are calculated using the following formula: Here, P represents the average value.

[0101] It should be noted that see Figure 5 -d, the discrete eigenvalue reflects the fluctuation intensity of the entire leakage pressure curve. If the discrete eigenvalue is large, it means that the entire curve is unstable and the fluctuation intensity is large. Training the model based on the discrete eigenvalue can enable the model to identify the fluctuation intensity and stability characteristics of the monitoring curve, thereby obtaining more accurate early warning data verification.

[0102] S24. Generate a feature training matrix based on the similarity eigenvalues, the linear fitting eigenvalues, the fluctuation eigenvalues ​​and the discrete eigenvalues.

[0103] Among them, the feature training matrix is ​​the final data input into the model for model training.

[0104] In some embodiments, step S24 (generating a feature training matrix based on the similarity eigenvalue, the linear fit eigenvalue, the fluctuation eigenvalue, and the discrete eigenvalue) specifically includes steps S241-S243: S241. Construct a leakage feature matrix based on similar eigenvalues, linear fitting eigenvalues, fluctuation eigenvalues ​​and discrete eigenvalues ​​corresponding to the historical training curve of the leakage state.

[0105] The leakage feature matrix represents a eigenvalue matrix corresponding to the historical training curve under the leakage state, and the model can be trained to identify typical features of the leakage state.

[0106]

[0107] S242: construct a non-leakage feature matrix based on similar eigenvalues, linear fitting eigenvalues, fluctuation eigenvalues ​​and discrete eigenvalues ​​corresponding to the historical training curve of the non-leakage state.

[0108]

[0109] It should be noted that the non-leakage feature matrix represents a eigenvalue matrix corresponding to the historical training curve under the non-leakage state, and the model can be trained to recognize typical features in the non-leakage state.

[0110] S243: stack the leakage feature matrix and the non-leakage feature matrix to generate a feature training matrix.

[0111] ; ;Pick is a 1×r column vector with a value of 1, which is used as a leak indicator. It is a 1×q column vector with the value -1 as a non-leakage marker.

[0112] It should be noted that stacking refers to merging the leakage feature matrix and the non-leakage feature matrix row by row to generate a complete matrix containing all state features, namely, the feature training matrix. Through stacking, the model can simultaneously obtain all features in the leakage and non-leakage states as unified input data.

[0113] S25. Construct a state label matrix based on the feature training matrix and each state classification label, wherein the state classification labels of the state label matrix correspond to the order of the historical training curves in the feature training matrix.

[0114] Among them, the state label matrix is ​​a matrix composed of state classification labels. Each row corresponds to a historical training curve in the feature training matrix. The label represents the state of the curve. The matrix is ​​used to supervise the training of the model. Through the state label matrix, the model can learn to distinguish between leaky and non-leaky states based on the features in the feature training matrix.

[0115] ; ; Among them, the number of leaks is l.

[0116] S26. Train the initial verification model based on the feature training matrix and the state label matrix to generate an early warning verification model.

[0117] It should be noted that the bidirectional long short-term memory recurrent neural network is trained through the feature training matrix and the state label matrix, so that it gradually learns the association between features and states, and generates a trained bidirectional long short-term memory recurrent neural network, that is, a warning verification model.

[0118] It should be noted that the bidirectional long short-term memory recurrent neural network (BiLSTM) is an extended neural network model based on the long short-term memory unit (LSTM), which has a strong ability to process time series data. Unlike the traditional LSTM, BiLSTM can simultaneously process the forward (past) and backward (future) information of the sequence, so it can capture contextual information more comprehensively in time-dependent tasks.

[0119] The LSTM unit consists of several key parts, including the forget gate, input gate, unit state update and output gate. The forget gate is responsible for filtering past information that needs to be forgotten, the input gate determines which new information needs to be stored in the unit state, and finally determines the change of the current state through the unit state update. The output gate generates the final output based on the unit state.

[0120] In BiLSTM, the forward LSTM processes the sequence from left to right to capture the past information of the input sequence, while the backward LSTM processes the same sequence from right to left to capture future information. By merging the outputs of the forward and backward LSTMs, BiLSTM can simultaneously consider the information before and after at each time point to obtain the complete context. The bidirectional processing method gives BiLSTM a clear advantage in processing tasks with strong time dependencies.

[0121] In cloud computing, BiLSTM can be used to analyze time series data such as pressure curves. Through forward and backward bidirectional calculations, BiLSTM can make comprehensive judgments on the data before and after the warning moment of the edge gateway, provide more accurate predictions and analysis, and improve the accuracy and reliability of warning data.

[0122] S3. Receive the pipe network early warning data obtained by the edge gateway based on the monitoring configuration parameters, send the pipe network early warning data to the early warning verification model for verification, and obtain early warning time series data, wherein the early warning time series data includes preceding data and succeeding data.

[0123] Among them, the pipeline network early warning data represents the key data sent to the cloud after the edge gateway performs preliminary processing on the monitored data, such as the location of the gas pipeline network initially identified as leaking and the corresponding key monitoring data; the early warning time series data represents the time series information input when the early warning verification model processes the pipeline network early warning data, which describes the changes of the monitoring values ​​over time; the preceding data represents the changes of the monitoring values ​​over time before the early warning moment; the following data represents the changes of the monitoring values ​​over time after the early warning moment.

[0124] It should be noted that the previous data helps the analysis model understand past trends, and they are an important basis for judging the current status. The subsequent data is used to monitor the status changes after the warning occurs, so as to continuously evaluate the severity of the leak or whether it is a missed report, false alarm, etc. The early warning verification model can provide continuous feature analysis, so as to better judge the missed report caused by small-scale or slow leaks and the false alarm caused by external interference, and improve the accuracy and reliability of the early warning data.

[0125] S4. Obtaining an early warning verification result based on the preceding data and the succeeding data, and when the early warning verification result is a leak, generating a leak early warning interface, and sending the leak early warning interface to a maintenance terminal.

[0126] Among them, the early warning verification result represents the final judgment result obtained by the early warning verification model, which shows whether the early warning data uploaded by the edge terminal is accurate. If the verification result obtained in the cloud also shows that the gas pipeline network is leaking, it means that the pipeline network does have a leak or other abnormal conditions; the leakage early warning interface represents a visual interface for displaying leakage early warning information to maintenance personnel, which may include information such as the time, location, possible causes, etc. of the leakage, which can be determined according to actual conditions; the maintenance end represents the terminal equipment used by maintenance personnel, such as mobile phones, computers, etc.

[0127] Specifically, the cloud verifies the warning data sent by the edge gateway, classifies the warning data uploaded by the edge gateway, marks the warning data of false alarms, does not perform warning operations and feeds back the verification results to the edge gateway. For truly missed warning data, the cloud generates a warning interface based on the warning verification results and sends it to the maintenance personnel so that gas leaks can be dealt with in a timely manner.

[0128] In some embodiments, step S4 (obtaining a warning verification result based on the preceding data and the succeeding data) specifically includes steps S41-S42: S41. Perform feature extraction based on the preceding data and the succeeding data to obtain preceding feature values ​​and succeeding feature values ​​respectively.

[0129] The preceding characteristic value represents the characteristic value of the pipe network monitoring data before the warning moment; and the subsequent characteristic value represents the characteristic value of the pipe network monitoring data after the warning moment.

[0130] S42: Integrate the preceding feature value and the succeeding feature value to obtain a warning association value, and compare the warning association value with the training feature value to obtain a warning verification result.

[0131] Among them, the warning association value is a numerical value or feature combination generated by integrating the previous feature value and the subsequent feature value. By comparing it with the historical training feature value, it can be determined whether the current state matches the training feature value of the historical leakage event and whether the warning data uploaded by the edge gateway is accurate.

[0132] It should be noted that integration refers to combining the preceding characteristic value with the succeeding characteristic value to obtain a complete monitoring characteristic value. The warning correlation value can be used to compare the training characteristic value to determine whether the current status of the pipeline network is a leak. If the match is high, the system will determine it as a leak. The warning data uploaded by the edge gateway is accurate, and a leakage warning interface is generated to notify maintenance personnel so that they can take maintenance measures in time.

[0133] In this embodiment, a gas pipeline leakage warning method based on edge-cloud collaboration is provided, including: responding to leakage monitoring instructions, obtaining pipeline operation data, generating monitoring configuration parameters based on the pipeline operation data, and sending the monitoring configuration parameters to the edge gateway; receiving a historical training curve library, performing feature extraction on each historical training curve in the historical training curve library, obtaining training feature values, training an initial verification model based on the training feature values, and generating a warning verification model; receiving pipeline warning data obtained by the edge gateway based on the monitoring configuration parameters, sending the pipeline warning data to the warning verification model for verification, and obtaining warning time series data, wherein the warning time series data includes preceding data and succeeding data; obtaining a warning verification result based on the preceding data and the succeeding data, and when the warning verification result is a leak, generating a leakage warning interface, and sending the leakage warning interface to the maintenance end. The present invention realizes efficient early warning of gas pipeline leakage by combining edge computing with cloud computing. After the edge device performs real-time preliminary processing and analysis on the pipeline data, the pipeline warning data with reduced noise and redundant information is uploaded to the cloud. The cloud uses powerful computing and storage capabilities to further analyze the pipeline warning data in depth to obtain early warning verification results. When the early warning verification result is a leak, a leakage early warning interface is generated and sent to the maintenance end, thereby improving the accuracy of pipeline leakage early warning. The cloud adopts an early warning verification model in the analysis process, which can accurately process time series data and discover minor anomalies, reducing underreporting caused by small-scale or slow leaks and false alarms caused by external interference, thereby improving the efficiency and reliability of pipeline leakage early warning.

[0134] like Figure 6 FIG. 1 is a schematic diagram of a gas pipeline network leakage warning system based on edge-cloud collaboration provided by an embodiment of the present invention. The gas pipeline network leakage warning system based on edge-cloud collaboration includes: The parameter configuration module is used to respond to the leakage monitoring instruction, obtain the pipe network operation data, generate the monitoring configuration parameters based on the pipe network operation data, and send the monitoring configuration parameters to the edge gateway.

[0135] The model acquisition module is used to receive a historical training curve library, extract features from each historical warning curve in the historical training curve library, obtain training feature values, train an initial verification model based on the training feature values, and generate a warning verification model.

[0136] The timing verification module is used to receive the pipe network warning data obtained by the edge gateway based on the monitoring configuration parameters, send the pipe network warning data to the warning verification model for verification, and obtain warning time series data, which includes preceding data and succeeding data.

[0137] The interface generation module is used to obtain the early warning verification result based on the preceding data and the following data, and when the early warning verification result is a leak, generate a leak early warning interface, and send the leak early warning interface to the maintenance end.

[0138] It should be noted that the above-mentioned gas pipeline leakage warning system based on edge-cloud collaboration can execute the gas pipeline leakage warning method based on edge-cloud collaboration provided in the embodiment of the present invention. For technical details and beneficial effects that are not fully described in the embodiment of the gas pipeline leakage warning system based on edge-cloud collaboration, please refer to the gas pipeline leakage warning method based on edge-cloud collaboration provided in the embodiment of the present invention.

[0139] Through the description of the above embodiments, it can be clearly understood by those skilled in the art that each embodiment can be implemented by means of software plus a general hardware platform, or by hardware. It can be understood by those skilled in the art that all or part of the processes in the above embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above methods. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0140] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them. Under the concept of the present invention, the technical features in the above embodiments or different embodiments may also be combined, the steps may be implemented in any order, and there are many other changes in different aspects of the present invention as described above, which are not provided in detail for the sake of simplicity. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person of ordinary skill in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A gas pipeline network leakage early warning method based on edge-cloud collaboration, characterized in that: The method comprises: Responding to a leakage monitoring instruction, obtaining pipe network operation data, generating monitoring configuration parameters based on the pipe network operation data, and sending the monitoring configuration parameters to an edge gateway; Receiving a historical training curve library, performing feature extraction on each historical training curve in the historical training curve library, obtaining a training feature value, training an initial verification model based on the training feature value, and generating an early warning verification model; Receiving the pipe network early warning data obtained by the edge gateway based on the monitoring configuration parameters, sending the pipe network early warning data to the early warning verification model for verification, and obtaining early warning time series data, wherein the early warning time series data includes pre-order data and post-order data; A warning verification result is obtained based on the preceding data and the succeeding data, and when the warning verification result is a leak, a leak warning interface is generated, and the leak warning interface is sent to a maintenance terminal.

2. The gas pipeline network leakage early warning method based on edge-cloud collaboration according to claim 1 is characterized in that: Responding to a leakage monitoring instruction, obtaining pipe network operation data, generating monitoring configuration parameters based on the pipe network operation data, and sending the monitoring configuration parameters to an edge gateway, including: In response to the leakage monitoring instruction, receiving the pipe network operation data sent by the edge gateway, the pipe network operation data including point data, pipeline data and environmental interference data; Retrieving preset parameter configuration strategies to identify the pipeline data, obtaining transportation attribute data and historical monitoring data, and determining basic configuration parameters according to the historical monitoring data; Based on the point data, regional terrain data, land use data and energy supply depth data are obtained, and a preset parameter impact assessment table is retrieved to respectively obtain point assessment correlation values ​​between the transportation attribute data and the regional terrain data, the land use data and the energy supply depth data, and a point correction value is determined based on each point assessment correlation value; Acquire meteorological condition data and traffic flow data based on the environmental interference data, acquire environmental assessment association values ​​between the transportation attribute data and the meteorological condition data and the traffic flow data respectively according to the preset parameter impact assessment table, and determine an environmental correction value based on each environmental assessment association value; Generate monitoring configuration parameters based on the point correction value, the environment correction value and the basic configuration parameters, and send the monitoring configuration parameters to the edge gateway.

3. The gas pipeline network leakage early warning method based on edge-cloud collaboration according to claim 1 is characterized in that: Receiving a historical training curve library, performing feature extraction on each historical training curve in the historical training curve library, obtaining a training feature value, training an initial verification model based on the training feature value, and generating an early warning verification model, including: Responding to the model training instruction, receiving the historical training curve library, identifying the historical training database, and obtaining the state classification label of each historical training curve; Determine a labeling state of a corresponding historical training curve based on the state classification label, wherein the labeling state includes a leakage state and a non-leakage state; Performing feature extraction on the historical training curves of the leakage state and the non-leakage state respectively to obtain corresponding training feature values, wherein the training feature values ​​include similarity feature values, linear fitting feature values, fluctuation feature values ​​and discrete feature values; Generate a feature training matrix based on the similarity eigenvalue, the linear fitting eigenvalue, the fluctuation eigenvalue and the discrete eigenvalue; Constructing a state label matrix based on the feature training matrix and each state classification label, wherein the state classification labels of the state label matrix correspond to the order of the historical training curves in the feature training matrix; The initial verification model is trained based on the feature training matrix and the state label matrix to generate an early warning verification model.

4. The gas pipeline network leakage early warning method based on edge-cloud collaboration according to claim 3 is characterized in that: Generating a feature training matrix based on the similarity eigenvalue, the linear fitting eigenvalue, the fluctuation eigenvalue and the discrete eigenvalue includes: Constructing a leakage feature matrix based on similar eigenvalues, linear fitting eigenvalues, fluctuation eigenvalues ​​and discrete eigenvalues ​​corresponding to the historical training curve of the leakage state; Constructing a non-leakage feature matrix based on similar eigenvalues, linear fitting eigenvalues, fluctuation eigenvalues ​​and discrete eigenvalues ​​corresponding to the historical training curve of the non-leakage state; The leakage feature matrix and the non-leakage feature matrix are stacked to generate a feature training matrix.

5. The gas pipeline network leakage early warning method based on edge-cloud collaboration according to claim 4 is characterized in that: The historical training curve of the leakage state is a leakage pressure curve, and feature extraction is performed on the historical training curve of the leakage state to obtain a training feature value, including: Acquire each leakage pressure data point based on the leakage pressure curve, combine any two leakage pressure data points, and generate a pressure difference similarity analysis group; Recalling a preset similarity analysis function to obtain the pressure difference similarity value of each pressure difference similarity analysis group, and generating a pressure difference similarity matrix based on each pressure difference similarity value; A maximum pressure difference interval is obtained based on the pressure difference similarity matrix, and the similarity eigenvalue is determined according to the maximum pressure difference interval.

6. The gas pipeline network leakage early warning method based on edge-cloud collaboration according to claim 5 is characterized in that: Extracting features from each historical training curve in the historical training curve library to obtain training feature values ​​also includes: Identify the leakage pressure curve and obtain the warning time label corresponding to each leakage pressure data point; Recalling a preset linear regression model to fit each leakage pressure data point and the corresponding warning time label, obtaining a tilt coefficient and an initial offset, and determining the tilt coefficient as a tilt characteristic value; The initial offset is determined as an intercept eigenvalue, and a linear fitting eigenvalue is obtained based on the tilt eigenvalue and the intercept eigenvalue.

7. The gas pipeline network leakage early warning method based on edge-cloud collaboration according to claim 6 is characterized in that: Extracting features from each historical training curve in the historical training curve library to obtain training feature values ​​also includes: Calculate the mean of the leakage pressure values ​​corresponding to each leakage pressure data point to obtain the pressure average value, and locate the fluctuation baseline based on the pressure average value; Determine the positive deviation area and the negative deviation area based on the fluctuation baseline, and call a preset area solution strategy to obtain the positive deviation area of ​​the positive deviation area and the negative deviation area of ​​the negative deviation area; An area difference is obtained based on the positive bias area and the negative bias area, and the area difference is determined as a fluctuation characteristic value.

8. The gas pipeline network leakage early warning method based on edge-cloud collaboration according to claim 7 is characterized in that: Extracting features from each historical training curve in the historical training curve library to obtain training feature values ​​also includes: Calculate the difference between the pressure average value and the leakage pressure value corresponding to each leakage pressure data point to obtain the pressure difference fluctuation amount; The square of each pressure difference fluctuation amount is calculated to obtain the single-point fluctuation variance, and the cumulative fluctuation variance is obtained based on the sum of the single-point fluctuation variances, and the cumulative fluctuation variance is determined as a discrete eigenvalue.

9. The gas pipeline network leakage early warning method based on edge-cloud collaboration according to claim 8 is characterized in that: Obtaining a warning verification result based on the preceding data and the following data includes: Perform feature extraction based on the preceding data and the succeeding data to obtain a preceding feature value and a succeeding feature value respectively; The preceding feature value and the succeeding feature value are integrated to obtain a warning correlation value, and the warning correlation value is compared with the training feature value to obtain a warning verification result.

10. A gas pipeline network leakage early warning system based on edge-cloud collaboration, characterized in that: The system comprises: A parameter configuration module, used to respond to a leakage monitoring instruction, obtain pipe network operation data, generate monitoring configuration parameters based on the pipe network operation data, and send the monitoring configuration parameters to an edge gateway; A model acquisition module is used to receive a historical training curve library, perform feature extraction on each historical warning curve in the historical training curve library, obtain training feature values, train an initial verification model based on the training feature values, and generate a warning verification model; A timing verification module, used to receive the pipe network warning data obtained by the edge gateway based on the monitoring configuration parameters, send the pipe network warning data to the warning verification model for verification, and obtain warning time series data, wherein the warning time series data includes pre-sequence data and post-sequence data; The interface generation module is used to obtain the early warning verification result based on the preceding data and the following data, and when the early warning verification result is a leak, generate a leak early warning interface, and send the leak early warning interface to the maintenance end.

Images