Intelligent gas pipeline risk monitoring method, medium, equipment and system

Abstract

The invention relates to an intelligent gas pipeline risk monitoring method, medium, equipment and system.The intelligent gas pipeline risk monitoring method comprises the steps that a gas emergency cut-off valve and a multi-dimensional information collecting device are arranged in a gas pipeline network, real-time multi-modal data of the gas pipeline network are collected, and after preprocessing, associated characteristic quantities are extracted based on an edge calculation analysis method; a multi-model learning discriminator is constructed, and the operation state and the fault occurrence probability of the gas pipeline network are predicted based on the associated characteristic quantity; media and equipment are realized based on the method. The system comprises one or more gas emergency cut-off valves arranged in a gas pipeline network in a distributed manner and one or more multi-dimensional information acquisition devices, and is used for acquiring real-time multi-modal data in a gas pipeline, predicting the running state of the gas pipeline network and the probability of failure, and controlling the opening and closing of the gas emergency cut-off valves. Comprising but not limited to timely pipe network fault positioning, processing and early warning, and accidents such as leakage or explosion of the gas pipe network in the conveying process or the gas using process of a user are avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic digital data processing, and in particular to a method, medium, equipment and system for intelligent monitoring of gas pipeline risks. Background Art

[0002] The electromagnetic gas emergency shut-off valve is a safety emergency shut-off device for gas pipelines, which can be connected to the gas leak alarm system or to the fire protection and other intelligent alarm control terminal modules. When a gas leak or other accident occurs, the alarm sounds, the gas emergency shut-off valve is energized, and the valve automatically closes, achieving a remote and rapid response on site, automatically cutting off the gas supply source, ensuring the safety of gas transmission and gas use in the gas pipeline, and avoiding safety accidents such as fire and explosion caused by gas pipeline leakage.

[0003] However, leakage and explosion failures in gas pipelines and networks are often caused by a single factor, which is usually caused by the coupling of multiple factors. The electromagnetic gas emergency shut-off valve in the existing technology is subject to environmental interference such as unstable power supply and strong external vibrations, and there is a possibility that the gas supply cannot be cut off in time, posing a threat to the safety of gas transportation. Existing gas pipeline safety diagnosis methods are almost all carried out for a single failure mode, and cannot effectively and reliably trace the cause of the failure. In addition, the electromagnetic gas emergency shut-off valve cannot automatically detect the multi-dimensional characteristic quantities in the gas pipeline, including but not limited to pressure, stress, gas concentration, temperature and flow, etc., and can only be activated when the electrical system fails. Therefore, for some long-term unattended gas transportation occasions, other technologies need to be introduced to monitor and maintain the equipment operation status to ensure the safe operation of the gas pipeline network.

[0004] With the rapid development of artificial intelligence and Internet of Things technology, on the basis of the existing electromagnetic gas emergency shut-off valve manufacturing and processing technology, artificial intelligence and Internet of Things technology are integrated to realize real-time monitoring of characteristic quantities such as gas pressure, stress, gas concentration, temperature and flow in the gas pipeline, and make early warning judgments, which can make the electromagnetic gas emergency shut-off valve have more intelligent functions; further establishment of an Internet of Things system for gas pipeline monitoring based on electromagnetic gas emergency shut-off valve can monitor the operating status of each monitoring node of the gas pipeline network in real time, which will be of great significance in ensuring the safe transportation of gas, the safety of users' life and property, and maintaining the harmony and stability of the city. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a method, medium, equipment and system for intelligent monitoring of gas pipeline risks.

[0006] The technical solution adopted by the present invention is a method for intelligent monitoring of gas pipeline risks. The method sets a gas emergency shut-off valve and a multidimensional information acquisition device in the gas pipeline network. The multidimensional information acquisition device collects real-time multimodal data of the gas pipeline network. After preprocessing the multimodal data, the associated feature quantities of the gas pipeline network are extracted based on the edge computing analysis method; a multi-model learning discriminator is constructed to predict the operating status of the gas pipeline network and the probability of failure based on the associated feature quantities.

[0007] Preferably, the multi-dimensional information collection device collects the pressure data of each monitoring node k in the gas pipeline network at time t Stress data Gas concentration data Temperature data and gas flow data Get real-time multimodal data satisfy k is a positive integer, and t is a non-negative integer.

[0008] Preferably, the multimodal data is preprocessed by denoising the obtained original real-time multimodal data and decomposing it into eigenmodal component data.

[0009] Preferably, decomposing the eigenmode component data comprises the following steps:

[0010] S1.1 Obtaining the original data in the continuous time domain T Extract the maximum and minimum values ​​and fit the original data based on the Hermite interpolation method The upper envelope, lower envelope and envelope coverage mean of Note the original signal The intermediate signal is satisfy

[0011]

[0012] S1.2 In the continuous time domain T, if The number of local minima and the number of zero crossings of the data are the same, and the upper and lower envelopes of the upper and lower envelopes have a mean of zero. If this intermediate signal is the first eigenmode component, otherwise repeat equation (1) until it is satisfied, and the eigenmode component is recorded as

[0013] S1.3 Filter the first eigenmode data from the original signal, denoted as

[0014]

[0015] Will As the original signal of the first decomposition, repeat n times until the intrinsic mode component function cannot be decomposed. The original signal is decomposed into n intrinsic mode data and residual terms Satisfying formula (3),

[0016]

[0017] S1.4 adds white noise to the intrinsic mode data to change the extreme point distribution of the original signal, and performs smoothing processing.

[0018] Preferably, S1.4 comprises the following steps:

[0019] S1.4.1 In the original signal Add white noise Get the signal Satisfying formula (4),

[0020]

[0021] Where, j = 1, 2, ..., m, j is the number of experiments adding white noise;

[0022] S1.4.2 Pair Solve the eigenmode components according to equations (1) to (3) to obtain the oth eigenmode component of the jth test: satisfy

[0023]

[0024] in, is the i-th eigenmode component of the j-th test, is the residual of the j-th test, and N is the number of times the j-th eigenmode test is performed;

[0025] S1.4.3 Calculate the mean of the eigencomponent set obtained from each decomposition, satisfying equation (6):

[0026]

[0027] Preferably, extracting the associated feature quantity of the gas pipeline network based on the edge computing analysis method includes the following steps:

[0028] S2.1 performs Hilbert transform on the eigenmode components to satisfy equation (7):

[0029]

[0030] Where ρ is the Cauchy principal quantity of the generalized integral, is the i-th eigenmode component at time τ;

[0031] S2.2 Calculate the analytical function Z(T) to satisfy equation (8),

[0032]

[0033] in, is the complex component in the analytical function, a(T) is the amplitude function, θ(T) is the phase function, which satisfy equations (9) and (10) respectively.

[0034]

[0035]

[0036] S2.3 obtains the time domain spectrum characteristics of the data set, which satisfies formula (11),

[0037]

[0038] Preferably, constructing the multi-model learning discriminator comprises the following steps:

[0039] S3.1 Select positive samples and negative samples to establish training sets (I1,y1),...,(I n ,y n ), where I n Represents the time domain spectrum characteristics of the input, y i are labels for positive and negative samples, with values ​​of 0 or 1;

[0040] S3.2 extracts the associated features of positive samples and negative samples based on edge computing analysis methods and establishes a feature vector set;

[0041] S3.3 Initialize the weights ζ corresponding to positive samples and negative samples respectively 1,i , normalize the weights;

[0042] S3.4 For each feature I i Train a classifier whose error rate is the weight ζ of the sample distribution t Measure, t Satisfying formula (13),

[0043] ζ t =∑ i ζ i |h j (I i )-y i | (13)

[0044] S3.5 Single Learning ν j (x) satisfies formula (14),

[0045]

[0046] Among them, f j is a feature (generally refers to the data features extracted above), θ j is the threshold, p j is the bias position, which is a parameter of the classifier obtained by multiple learning and training of multiple single features in the above single learning formula, including the threshold, which is obtained through self-training based on the data features of each category; single learning here refers to learning and classifying a certain category of features, because the data collected above is multi-modal and multi-category data, first single learning is performed on one category of data, and then training and learning of multi-model data is realized;

[0047] Sort the error rates of the classifiers and take the classifier with the lowest error rate ν j , whose error rate is ε t , update the weights When sample I i When correctly classified, e i =0, otherwise e i =1,β t Satisfying formula (15),

[0048]

[0049] S3.6 builds a multi-model learning discriminator that satisfies equation (16),

[0050]

[0051] in,

[0052] A computer-readable storage medium stores a gas pipeline risk intelligent monitoring program, which implements the above-mentioned gas pipeline risk intelligent monitoring method when executed by a processor.

[0053] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the program, the above-mentioned intelligent monitoring method for gas pipeline risks is implemented.

[0054] A gas pipeline risk intelligent monitoring system, the system comprising:

[0055] One or more gas emergency shut-off valves are distributed in the gas pipeline network;

[0056] One or more multi-dimensional information collection devices are distributed in the gas pipeline network and are used to collect real-time multi-modal data of the gas pipeline network;

[0057] A control end is associated with the multi-dimensional information collection device, and adopts the gas pipeline risk intelligent monitoring method to dynamically monitor the gas pipeline network, predict the operating status of the gas pipeline network and the probability of failure, and control the opening and closing of the gas emergency shut-off valve.

[0058] The present invention provides a method, medium, equipment and system for intelligent monitoring of gas pipeline risks. The method sets a gas emergency shut-off valve and a multi-dimensional information collection device in a gas pipeline network, collects real-time multi-modal data of the gas pipeline network, and extracts the associated feature quantity of the gas pipeline network based on an edge computing analysis method after pre-processing the multi-modal data; constructs a multi-model learning discriminator, and predicts the operating status and the probability of failure of the gas pipeline network based on the associated feature quantity; implements the medium and equipment based on the method; the system includes one or more gas emergency shut-off valves and one or more multi-dimensional information collection devices distributed in the gas pipeline network, realizes real-time multi-modal data collection in the gas pipeline, predicts the operating status and the probability of failure of the gas pipeline network, and controls the opening and closing of the gas emergency shut-off valve, including but not limited to timely positioning, processing and early warning of pipeline network faults, to avoid accidents such as leakage or explosion of the gas pipeline network during the transportation process or the user's gas use process.

[0059] The beneficial effects of the present invention are:

[0060] (1) Using edge computing and artificial intelligence technologies, we can achieve real-time monitoring of gas in gas pipelines, analyze and process information such as gas flow, temperature, and pressure, so as to predict the probability of gas leakage, and provide timely warning and processing to avoid gas accidents;

[0061] (2) Build a system, create cloud servers and intelligent Internet of Things technology, support the access of more monitoring equipment and data collection, analysis and transmission, and feed back data to the control end in real time, so that the system terminal can remotely control the opening and closing of the shut-off valve to avoid safety accidents such as fire and explosion caused by gas leakage;

[0062] (3) It can realize dynamic monitoring of the gas pipeline operation status and fault probability calculation, and realize accident warning. It is of great significance in the safety monitoring and fault assessment and prediction of gas transportation and use, and facilitates the intelligent application expansion of electromagnetic gas emergency shut-off valves. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 is a flow chart of the method of the present invention;

[0064] Figure 2 It is a schematic diagram of the system structure of the present invention;

[0065] Figure 3 It is a schematic diagram of the application of the present invention. DETAILED DESCRIPTION

[0066] The present invention is further described in detail below in conjunction with embodiments, but the protection scope of the present invention is not limited thereto.

[0067] The present invention relates to a method for intelligently monitoring gas pipeline risks. The method comprises the following steps: arranging a gas emergency shut-off valve and a multi-dimensional information acquisition device in a gas pipeline network. The multi-dimensional information acquisition device acquires real-time multi-modal data of the gas pipeline network. After pre-processing the multi-modal data, the associated feature quantity of the gas pipeline network is extracted based on an edge computing analysis method. A multi-model learning discriminator is constructed to predict the operating state of the gas pipeline network and the probability of failure based on the associated feature quantity.

[0068] In the present invention, by collecting and analyzing characteristic quantities in the gas pipeline, the parameters of the electromagnetic emergency shut-off valve monitoring node corresponding to the gas pipeline are monitored in real time, including but not limited to pressure, stress, gas concentration, temperature, and flow, etc., and through the design of artificial intelligence algorithm, calculation and analysis are performed at the edge end, and the data collected by the gas Internet of Things node is monitored and analyzed from multi-dimensional information to complete the prediction and early warning of the probability of accidents such as leakage and explosion in the gas pipeline, and the prediction results are fed back to the control terminal. By remotely controlling and operating the solenoid valve system of the pipeline node in advance, the gas valve is emergency cut or closed to ensure that the gas is more intelligent and safe during transmission and user use.

[0069] The method is explained below in combination with specific content.

[0070] The multi-dimensional information collection device collects the pressure data of each monitoring node k in the gas pipeline network at time t Stress data Gas concentration data Temperature data and gas flow data Get real-time multimodal data satisfy k is a positive integer, and t is a non-negative integer.

[0071] The multimodal data is preprocessed to denoise the original real-time multimodal data and decompose it into intrinsic mode component data.

[0072] In the present invention, by arranging and installing multi-dimensional information collection devices including but not limited to pressure sensors, stress sensors, gas concentration detectors, temperature sensors and gas flow meters on the solenoid valve, the collection, acquisition and risk failure analysis of common data of the gas pipeline are realized; the pressure sensor, stress sensor, gas concentration detector, temperature sensor and gas flow meter can be set separately or integrated, which is easy for technical personnel in this field to understand. The integrated multi-dimensional information collection device can more objectively reflect the status of a certain monitoring node in the gas pipeline network.

[0073] In the present invention, the pressure data of each monitoring node k at time t is Stress data Gas concentration data Temperature data and gas flow data Conduct analysis to achieve online monitoring and real-time early warning of the operating status of the gas pipeline;

[0074] Before analysis, the data needs to be preprocessed. Since the data obtained at the monitoring node is time series data, that is, time sampling of multidimensional data, and in actual operation, the monitoring node is affected by factors such as pipe or valve wear, corrosion or deformation and external vibration, the acquired data basically contains nonlinear and non-stationary noise, so it is necessary to denoise the sampled data at the edge first, retain the intrinsic data, and then extract better data features.

[0075] In the present invention, the collected signal data is divided into three parts: noise data, intrinsic data and time drift data. The collected original data is denoised according to the time scale characteristics, the original data is decomposed into a group of intrinsic mode component data, and the spectrum form of the data in the time domain is obtained through Hilbert transform, and the eigenvalue of the data is further obtained.

[0076] Specifically, decomposing the intrinsic mode component data includes the following steps:

[0077] S1.1 Obtaining the original data in the continuous time domain T Calculate the maximum and minimum values, consider the time continuity of the data in the continuous time domain, and fit the original data based on the Hermite interpolation method The upper envelope, lower envelope and envelope coverage mean of The intermediate signal of the original signal is satisfy

[0078]

[0079] S1.2 In the continuous time domain T, if The number of local minima and the number of zero crossings of the data are the same, and the upper and lower envelopes of the upper and lower envelopes have a mean of zero. If this intermediate signal is the first eigenmode component, otherwise repeat equation (1) until it is satisfied, and the eigenmode component is recorded as

[0080] S1.3 Filter the first eigenmode data from the original signal, denoted as

[0081]

[0082] Will As the original signal of the first decomposition, repeat the above operation for n times until the intrinsic mode component function cannot be decomposed. The original signal is decomposed into n intrinsic mode data and residual terms Satisfying formula (3),

[0083]

[0084] S1.4 adds white noise to the intrinsic mode data to change the extreme point distribution of the original signal, and performs smoothing processing.

[0085] In the present invention, since the original signal has discontinuous extreme points and multiple signals with small frequency differences, the above-mentioned intrinsic mode components will have "overshoot" or "undershoot" phenomena, resulting in modal aliasing in the original multidimensional component data. It is necessary to add white noise to the above-mentioned intrinsic mode data, change the extreme point distribution of the original signal, and perform smoothing processing to facilitate subsequent training. Specifically, the following steps are included:

[0086] S1.4.1 In the original signal Add white noise Get the signal Satisfying formula (4),

[0087]

[0088] Where, j = 1, 2, ..., m, m is the number of experiments adding white noise;

[0089] S1.4.2 Pair Solve the eigenmode components according to equations (1) to (3) to obtain the i-th eigenmode component of the j-th test: satisfy

[0090]

[0091] in, is the i-th eigenmode component of the j-th test, is the residual of the jth test, N is the number of eigenmodes of the jth test;

[0092] S1.4.3 Calculate the mean of the eigencomponent set obtained from each decomposition, satisfying equation (6):

[0093]

[0094] After completing the data preprocessing, the above intrinsic mode component set is subjected to Hilbert to obtain the time domain spectrum characteristics of the original data; specifically, the extraction of the associated feature quantity of the gas pipeline network based on the edge computing analysis method includes the following steps:

[0095] S2.1 performs Hilbert transform on the eigenmode components to satisfy equation (7):

[0096]

[0097] Where ρ is the Cauchy principal quantity of the generalized integral, is the i-th eigenmode component at time τ;

[0098] S2.2 Calculate the analytical function Z(T) to satisfy equation (8),

[0099]

[0100] in, is the complex component in the analytical function, a(T) is the amplitude function, θ(T) is the phase function, which satisfy equations (9) and (10) respectively.

[0101]

[0102] S2.3 obtains the time domain spectrum characteristics of the data set, which satisfies formula (11),

[0103]

[0104] Taking into account that the data collected on-site are multidimensional data and the features of each type are interrelated, the feature representation ability is not strong enough when using a single feature model, and the model learning efficiency is low. Therefore, a multi-model learning discriminator is used to combine multiple single models through an integration strategy to weaken the preference for a single learning model, so that multiple single models can learn features better and more comprehensively, which can not only give play to the advantages of each model, but also avoid the limitations of a single model.

[0105] Specifically, constructing the multi-model learning discriminator includes the following steps:

[0106] S3.1 selects positive samples and negative samples, which are generally multidimensional data collected from gas pipelines without fault warning and multidimensional data collected from gas pipelines with faults, and establishes training sets (I1, y1),..., (I n ,y n ), where I i Represents the time domain spectrum characteristics of the input, y i is the label of positive and negative samples, i is 0 or 1, generally speaking, y i =0, 1 represents the output is a negative sample (fault) and a positive sample (normal) respectively;

[0107] S3.2 For samples in the training set, extract the associated feature quantities of positive samples and negative samples based on the edge computing analysis method to establish a feature vector set;

[0108] S3.3 Initialize the weights ζ corresponding to positive samples and negative samples respectively 1,i , to be precise, corresponds to y i =0,1, initialize weight ζ 1,i for and Where m and l are the number of negative samples and positive samples respectively; the weights are normalized as shown in formula (12):

[0109]

[0110] S3.4 trains a classifier for each feature, and the error rate of the classifier is the weight ζ of the sample distribution t Measure, t Satisfying formula (13),

[0111] ζ t =∑ i ζ i |h j (x i )-y i | (13)

[0112] S3.5 Single Learning ν j (x) satisfies formula (14),

[0113]

[0114] Among them, f j is the feature, θ j is the threshold, p j is the bias position (indicating the direction of the inequality);

[0115] Sort the error rates of the classifiers and take the classifier with the lowest error rate ν t , whose error rate is ε t , update the weights When sample I i When correctly classified, e i =0, otherwise e i =1,β t Satisfying formula (15),

[0116]

[0117] S3.6 builds a multi-model learning discriminator that satisfies equation (16),

[0118]

[0119] in,

[0120] Through the above algorithm, it is possible to monitor the operating status of the gas pipeline network at the monitoring node in real time, and to provide early warning and real-time analysis of the probability of failure of the gas pipeline network at the node.

[0121] The present invention also relates to a computer-readable storage medium on which a gas pipeline risk intelligent monitoring program is stored. When the program is executed by a processor, the above-mentioned gas pipeline risk intelligent monitoring method is implemented.

[0122] The present invention also relates to a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the above-mentioned gas pipeline risk intelligent monitoring method is implemented.

[0123] The present invention also relates to a gas pipeline risk intelligent monitoring system, the system comprising:

[0124] One or more gas emergency shut-off valves are distributed in the gas pipeline network;

[0125] One or more multi-dimensional information collection devices are distributed in the gas pipeline network and are used to collect real-time multi-modal data of the gas pipeline network;

[0126] A control end is associated with the multi-dimensional information collection device, and adopts the gas pipeline risk intelligent monitoring method to dynamically monitor the gas pipeline network, predict the operating status of the gas pipeline network and the probability of failure, and control the opening and closing of the gas emergency shut-off valve.

[0127] In the present invention, supervised learning is performed on the multimodal feature data obtained from the emergency shut-off valve monitoring node in the gas pipe network system to realize online monitoring and real-time early warning of the operating status of the gas pipe network; supervised learning is adopted to learn and model the empirical data, and predict the data to realize the binary classification problem of data characterization; a gas pipeline risk intelligent monitoring system with the gas emergency shut-off valve as the monitoring node is established, and 5G technology is used to realize the remote transmission of the collected monitoring data and prediction results between the multidimensional information collection device and the control end, thereby completing the overall monitoring and control of the gas pipe network by the gas control terminal (control end), and avoiding accidents such as leakage or explosion in the gas pipe network during the transportation process or the user's gas use process.

[0128] In the present invention, an Internet of Things system platform for intelligent monitoring of gas pipeline risks can also be constructed during the application process. Through the data collected at the gas pipeline nodes, edge processing is performed on pressure, stress, gas concentration, flow and temperature, and an early warning is submitted at the platform, which is then fed back to the specific gas pipeline nodes for control.

[0129] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0130] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0131] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1A function specified in one or more boxes.

[0132] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0133] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0134] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A gas pipeline risk intelligent monitoring method, characterized by: The method sets a gas emergency shut-off valve and a multi-dimensional information acquisition device in the gas pipeline network, the multi-dimensional information acquisition device collects real-time multi-modal data of the gas pipeline network, and extracts the associated feature quantity of the gas pipeline network based on the edge computing analysis method after pre-processing the multi-modal data; A multi-model learning discriminator is constructed to predict the operating status and failure probability of the gas pipeline network based on the associated feature quantities.

2. A gas pipeline risk intelligent monitoring method according to claim 1, characterized in that: The multi-dimensional information collection device collects the pressure data of each monitoring node k in the gas pipeline network at time t Stress data Gas concentration data Temperature data and gas flow data Get real-time multimodal data satisfy k is a positive integer, and t is a non-negative integer.

3. A gas pipeline risk intelligent monitoring method according to claim 2, characterized in that: The multimodal data is preprocessed to denoise the original real-time multimodal data and decompose it into intrinsic mode component data.

4. A gas pipeline risk intelligent monitoring method according to claim 3, characterized in that: Decomposing the eigenmode component data includes the following steps: S1.1 Obtaining the original data in the continuous time domain T Extract the maximum and minimum values ​​and fit the original data based on the Hermite interpolation method The upper envelope, lower envelope and envelope coverage mean of The intermediate signal of the original signal is satisfy S1.2 In the continuous time domain T, if The number of local minima and the number of zero crossings of the data are the same, and the upper and lower envelopes of the upper and lower envelopes have a mean of zero. If this intermediate signal is the first eigenmode component, otherwise repeat equation (1) until it is satisfied, and the eigenmode component is recorded as S1.3 Filter the first eigenmode data from the original signal, denoted as Will As the original signal of the first decomposition, repeat n times until the intrinsic mode component function cannot be decomposed. The original signal is decomposed into n intrinsic mode data and residual terms Satisfying formula (3), S1.4 adds white noise to the intrinsic mode data to change the extreme point distribution of the original signal, and performs smoothing processing.

5. A gas pipeline risk intelligent monitoring method according to claim 4, characterized in that: S1.4 includes the following steps: S1.4.1 In the original signal Add white noise Get the signal Satisfying formula (4), Where, j = 1, 2, ..., m, j is the number of experiments adding white noise; S1.4.2 Pair Solve the eigenmode components according to equations (1) to (3) to obtain the oth eigenmode component of the jth test: satisfy in, is the i-th eigenmode component of the j-th test, is the residual of the jth test, N j is the number obtained by the jth eigenmode experiment; S1.4.3 Calculate the mean of the eigencomponent set obtained from each decomposition, satisfying equation (6):

6. A gas pipeline risk intelligent monitoring method according to claim 1, characterized in that: Extracting the associated feature quantities of the gas pipeline network based on the edge computing analysis method includes the following steps: S2.1 performs Hilbert transform on the eigenmode components to satisfy equation (7): Where ρ is the Cauchy principal quantity of the generalized integral, is the i-th eigenmode component at time τ; S2.2 Calculate the analytical function Z(T) to satisfy equation (8), in, is the complex component in the analytical function, a(T) is the amplitude function, θ(T) is the phase function, which satisfy equations (9) and (10) respectively. S2.3 obtains the time domain spectrum characteristics of the data set, which satisfies formula (11), 7. A gas pipeline risk intelligent monitoring method according to claim 1, characterized in that: Constructing the multi-model learning discriminator includes the following steps: S3.1 Select positive samples and negative samples to establish training sets (I1,y1),...,(I n ,y n ), where I n Represents the time domain spectrum characteristics of the input, y i are labels for positive and negative samples, with values ​​of 0 or 1; S3.2 extracts the associated features of positive samples and negative samples based on edge computing analysis methods and establishes a feature vector set; S3.3 Initialize the weights ζ corresponding to positive samples and negative samples respectively 1,i , normalize the weights; S3.4 For each feature I i Train a classifier whose error rate is the weight ζ of the sample distribution t Measure, t Satisfying formula (13), g t =∑ i g i |h j (I i )-y i | (13) S3.5 Single Learning ν j (x) satisfies formula (14), Among them, f j is the feature, θ j is the threshold, p j is the offset position; Sort the error rates of the classifiers and take the classifier with the lowest error rate ν j , whose error rate is ε t , update the weights When sample I i When correctly classified, e i =0, otherwise e i =1,β t Satisfying formula (15), S3.6 builds a multi-model learning discriminator that satisfies equation (16), in, 8. A computer-readable storage medium, characterized in that: A gas pipeline risk intelligent monitoring program is stored thereon, and when the program is executed by the processor, the gas pipeline risk intelligent monitoring method described in any one of claims 1 to 7 above is implemented.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the gas pipeline risk intelligent monitoring method described in any one of claims 1 to 7 is implemented.

10. A gas pipeline risk intelligent monitoring system, characterized by: The system comprises: one or more gas emergency shut-off valves, which are distributed in the gas pipeline network; One or more multi-dimensional information collection devices are distributed in the gas pipeline network and are used to collect real-time multi-modal data of the gas pipeline network; A control end is associated with the multi-dimensional information collection device, and adopts the gas pipeline risk intelligent monitoring method described in one of claims 1 to 7, which is used to dynamically monitor the gas pipeline network, predict the operating status of the gas pipeline network and the probability of failure, and control the opening and closing of the gas emergency shut-off valve.

Images