Intelligent gas valve with point laser methane sensor

Abstract

The application discloses a gas intelligent valve with a point laser methane sensor, and relates to the field of gas intelligent valves.The gas intelligent valve comprises a device installation module, a data acquisition module, a risk assessment module, a gas leakage positioning module, a valve self-adjusting module and a historical data analysis module.The device installation module is used for installing three point laser methane sensors to collect interface methane concentration values.The data acquisition module is used for acquiring pipeline parameters and environmental parameter data inside and outside the interface.The risk assessment module calculates a methane concentration change rate according to the methane concentration value, and assesses whether gas leakage exists at each interface.The gas leakage positioning module is used for constructing a gas leakage positioning model based on a neural network, and acquiring the probability of gas leakage risk at each interface.The valve self-adjusting module is used for confirming whether it is a leakage point according to the gas leakage risk probability.The historical data analysis module is used for calculating the precision rate, recall rate and F1-Score value by using historical data and on-site measured results.The gas pipeline leakage risk can be effectively diagnosed and positioned, and the valve opening degree can be intelligently controlled.

Description

Technical Field

[0001] This invention relates to the field of intelligent gas valves, specifically to an intelligent gas valve with a point-type laser methane sensor. Background Technology

[0002] The point-type laser methane sensor is a high-precision gas monitoring device based on tunable laser spectral absorption monitoring (TDLAS) technology. The laser source (such as a DFB laser) emits a laser of a specific wavelength that precisely matches the absorption peak of methane molecules (e.g., 1653.7 nm). When the laser passes through the gas to be measured, the methane molecules selectively absorb the laser energy of the specific wavelength, while the unabsorbed laser continues to propagate. The detector measures the intensity of the absorbed laser and, using Lambert-Beer's law, calculates the methane concentration in the gas. It also employs harmonic monitoring methods (such as second harmonic monitoring) and lock-in amplification technology to eliminate background gas interference and improve monitoring sensitivity, resulting in high precision and sensitivity.

[0003] As a critical infrastructure for energy supply and transmission, gas pipelines are prone to leaks due to aging, corrosion, and external damage during long-term operation. By precisely controlling the opening and closing of valves, large-scale gas leaks can be effectively suppressed, leak points can be located, leak risks can be reduced, gas utilization can be improved, and the safety and stability of energy supply and transmission can be guaranteed.

[0004] In existing technologies, gas leak early warning mostly relies on single concentration monitoring by point-type laser methane sensors, and risk is determined by fixed thresholds. However, this method is easily affected by parameters inside the pipeline (such as fluctuations in gas pressure and flow rate) and external environment (such as temperature, humidity, and dust interference), leading to concentration monitoring deviations, false alarms, and missed alarms. Frequent invalid early warnings can cause "early warning fatigue," reduce staff's trust in risk warnings and response efficiency, and make it difficult to meet the needs of gas safety management under complex working conditions. Summary of the Invention

[0005] To address the aforementioned technical issues, a smart gas valve with a point-type laser methane sensor is provided. This technical solution resolves the problems mentioned in the background technology, such as the susceptibility of concentration monitoring to internal pipeline parameters and external environment, leading to false alarms, missed alarms, and frequent invalid warnings causing "warning fatigue," reducing staff's trust in risk warnings and response efficiency, and making it difficult to adapt to the gas safety management needs under complex operating conditions.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A smart gas valve with a point-type laser methane sensor includes:

[0008] The equipment installation module is used to install three point-type laser methane sensors at a 120-degree angle in a ring at each interface of the gas transmission pipeline to collect methane concentration values ​​at different locations of the interface.

[0009] The data acquisition module is used to install acquisition devices both inside and outside each interface of the gas transmission pipeline to obtain pipeline parameters and environmental parameters inside and outside the interface.

[0010] The risk assessment module is used to assess whether there is a gas leak at each interface based on the methane concentration values ​​collected in real time by three point laser methane sensors and the calculated methane concentration change rate.

[0011] A gas leak location module is used to construct a neural network-based gas leak location model based on the data collected by the data acquisition module, and to obtain the probability of gas leak risk at each interface.

[0012] The valve self-adjustment module is used to prioritize adjusting the valve opening at both ends of the high-probability interface based on the probability of gas leakage risk, and combine the feedback results of the gas leakage location model based on neural network to confirm whether it is a leak point.

[0013] The historical data analysis module is used to calculate the precision, recall, and F1-Score of the gas leak location model based on neural networks using historical data and field measurement results, and to provide feedback on the stability and accuracy of the gas leak location model based on neural networks.

[0014] Preferably, the risk assessment module specifically includes:

[0015] A buffer window unit is used to maintain a circular buffer of length a for each sensor, storing the methane concentration values ​​collected the most recently a times. Each time new data arrives, the oldest value is discarded and the latest value is stored.

[0016] The rate of change unit is used to acquire methane monitoring concentration data from three point laser methane sensors simultaneously and calculate the absolute rate of change of methane concentration from the two most recent acquisitions.

[0017] A rate of change threshold unit is used to set a threshold for the absolute rate of change of methane concentration based on the statistical analysis of the historical absolute rate of change of methane concentration, using the mean and standard deviation of the rate of change over the past 24 hours.

[0018] The risk assessment unit is used to conduct risk assessments on each interface of the gas transmission pipeline based on the methane concentration value and the methane concentration change rate, using a primary screening and secondary verification approach.

[0019] A graded early warning unit is used to classify risk warnings into first-level warnings, second-level warnings, and warning suppression based on the judgment results of primary screening and secondary verification.

[0020] Preferably, the gas leak location module specifically includes:

[0021] Node attribute unit, which is used to construct the node attributes of the gas pipeline map based on the collected internal pipeline parameters and external environmental parameters;

[0022] The node attributes of the gas pipeline map include: external environmental parameters such as temperature and humidity, dust concentration, and air velocity, as well as internal environmental parameters such as gas pressure, velocity, pipeline inner diameter, cross-sectional area, and valve opening.

[0023] Edge attribute unit, the edge attribute unit is used to construct the edge attributes of the gas pipeline map based on the straight distance between interfaces, the elevation difference between interfaces, the pipeline connection angle and the flow direction identifier;

[0024] A pipeline mapping unit is used to construct a gas pipeline map based on the attributes of each interface node of the gas pipeline and the attributes of the edges with adjacent interfaces.

[0025] The data normalization unit, the pipeline map unit is used to normalize the parameter data, methane concentration and methane concentration change rate in the gas pipeline map based on the min-max normalization formula, so as to eliminate the influence of data dimensions.

[0026] The risk probability unit is used to construct a gas leak location model based on a neural network through graph convolutional layers, LSTM layers, and fully connected network layers, respectively. The dynamic safety threshold of methane concentration and the concentration change rate threshold output by the risk assessment module are used as input constraints of the model, and the output is a three-dimensional vector of interface number, risk probability, and leakage rate.

[0027] Preferably, the step of prioritizing the adjustment of valve openings at both ends of high-probability interfaces based on the probability of gas leakage risk, and confirming whether a leak point is identified by combining feedback results from a gas leak location model based on a neural network, specifically includes:

[0028] The interfaces are sorted from high to low leakage probability. The valves at both ends of the high-probability interfaces are finely adjusted at a fixed angle. The methane concentration changes of the interface and adjacent interfaces are collected based on a point-type laser methane sensor.

[0029] If the methane concentration decreases as the valves at both ends are adjusted, the interface is identified as a leak point, and the abnormal location is reported. If the methane concentration difference is within the specified range, the interface is identified as not a leak point, the valve opening is restored, and the above operation is repeated for the next highest probability interface until the leak point is located.

[0030] Preferably, the step of using historical data and field measurement results from the neural network-based gas leak location model to calculate the system's precision, recall, and F1-Score, and to provide feedback on the stability and accuracy of the neural network-based gas leak location model, specifically includes:

[0031] Simultaneously acquire monitoring data from a point-type laser methane sensor, judgment results from a neural network-based gas leak location model, and actual on-site monitoring results to form comprehensive historical data;

[0032] Based on comprehensive historical data, a historical data sample set is created, and the number of true positive samples, false positive samples, and false negative samples is determined by comparative analysis of the data.

[0033] The sample set must cover the typical and abnormal operating conditions of the system. The number of true positive samples refers to the number of samples that are actually positive and are predicted to be positive. The number of false positive samples refers to the number of samples that are actually negative but are predicted to be positive. The number of false negative samples refers to the number of samples that are actually positive but are predicted to be negative.

[0034] Precision is calculated based on the number of true positive samples and false positive samples, recall is calculated based on the number of true positive samples and false negative samples, and then the overall value is calculated using the F1-Score formula;

[0035] Based on the F1-Score value and the fluctuation range of its continuous calculation results, the stability and accuracy of the system risk verification and positioning are judged, and the performance status of the positioning model is fed back synchronously.

[0036] The F1-Score ranges from 0 to 1. The closer it is to 1, the better the overall performance of the system's risk verification and location. The smaller the fluctuation of the continuous calculation results, the more stable it is.

[0037] The performance status of the synchronous feedback positioning model refers to the comprehensive evaluation of the stability and accuracy of the positioning model based on the monitoring data of the point laser methane sensor, the judgment results of the positioning model, and the trend curve of the actual on-site monitoring results over the past 30 days, combined with the historical stability trend curve of F1-Score, and assisted by expert experience.

[0038] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0039] This invention provides a smart gas valve with a point-type laser methane sensor. Through the synergistic effect of a risk assessment module and a gas leak location module, it significantly improves the risk judgment and leak location capabilities of the point-type laser methane sensor in complex gas pipeline environments. The location module can deeply verify the risk assessment results, avoiding false alarms caused by gas diffusion at non-leak points and reducing "warning fatigue" caused by frequent invalid warnings. Based on this, the valve self-adjustment module dynamically adjusts the valve opening degree at both ends of the suspected leak point interface according to the judgment data, which not only reduces the leakage risk but also assists in accurate location through parameter changes during the adjustment process. Finally, by combining historical judgment data, field measurement results, and the F1-Score stability trend curve, the invention assists experts in evaluating the performance of the location model, achieving closed-loop optimization of the system. In summary, this invention comprehensively ensures the intelligent control capability and operational reliability of the smart gas valve for gas transmission. Attached Figure Description

[0040] Figure 1 This is a structural block diagram of a gas intelligent valve with a point-type laser methane sensor according to the present invention.

[0041] Figure 2 The present invention provides a flowchart for assessing the presence of gas leaks at each interface based on real-time methane concentration values ​​collected by three point-type laser methane sensors and the calculated methane concentration change rate.

[0042] Figure 3 The present invention constructs a gas leak location model based on a neural network based on the data collected by the data acquisition module, and obtains a probability flowchart of the gas leak risk at each interface. Detailed Implementation

[0043] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0044] Reference Figure 1 As shown, a smart gas valve with a point-type laser methane sensor includes:

[0045] The equipment installation module is used to install three point-type laser methane sensors at a 120-degree angle in a ring at each interface of the gas transmission pipeline to collect methane concentration values ​​at different locations of the interface.

[0046] The data acquisition module is used to install acquisition devices both inside and outside each interface of the gas transmission pipeline to obtain pipeline parameters and environmental parameters inside and outside the interface.

[0047] The risk assessment module is used to assess whether there is a gas leak at each interface based on the methane concentration values ​​collected in real time by three point laser methane sensors and the calculated methane concentration change rate.

[0048] A gas leak location module is used to construct a neural network-based gas leak location model based on the data collected by the data acquisition module, and to obtain the probability of gas leak risk at each interface.

[0049] The valve self-adjustment module is used to prioritize adjusting the valve opening at both ends of the high-probability interface based on the probability of gas leakage risk, and combine the feedback results of the gas leakage location model based on neural network to confirm whether it is a leak point.

[0050] The historical data analysis module is used to calculate the precision, recall, and F1-Score of the gas leak location model based on neural networks using historical data and field measurement results, and to provide feedback on the stability and accuracy of the gas leak location model based on neural networks.

[0051] Reference Figure 1 As shown, the risk assessment module specifically includes:

[0052] A buffer window unit is used to maintain a circular buffer of length a for each sensor, storing the methane concentration values ​​collected the most recently a times. Each time new data arrives, the oldest value is discarded and the latest value is stored.

[0053] The rate of change unit is used to acquire methane monitoring concentration data from three point laser methane sensors simultaneously and calculate the absolute rate of change of methane concentration from the two most recent acquisitions.

[0054] A rate of change threshold unit is used to set a threshold for the absolute rate of change of methane concentration based on the statistical analysis of the historical absolute rate of change of methane concentration, using the mean and standard deviation of the rate of change over the past 24 hours.

[0055] The risk assessment unit is used to conduct risk assessments on each interface of the gas transmission pipeline based on the methane concentration value and the methane concentration change rate, using a primary screening and secondary verification approach.

[0056] A graded early warning unit is used to classify risk warnings into first-level warnings, second-level warnings, and warning suppression based on the judgment results of primary screening and secondary verification.

[0057] Reference Figure 1 As shown, the gas leak location module specifically includes:

[0058] Node attribute unit, which is used to construct the node attributes of the gas pipeline map based on the collected internal pipeline parameters and external environmental parameters;

[0059] The node attributes of the gas pipeline map include: external environmental parameters such as temperature and humidity, dust concentration, and air velocity, as well as internal environmental parameters such as gas pressure, velocity, pipeline inner diameter, cross-sectional area, and valve opening.

[0060] Edge attribute unit, the edge attribute unit is used to construct the edge attributes of the gas pipeline map based on the straight distance between interfaces, the elevation difference between interfaces, the pipeline connection angle and the flow direction identifier;

[0061] A pipeline mapping unit is used to construct a gas pipeline map based on the attributes of each interface node of the gas pipeline and the attributes of the edges with adjacent interfaces.

[0062] The data normalization unit, the pipeline map unit is used to normalize the parameter data, methane concentration and methane concentration change rate in the gas pipeline map based on the min-max normalization formula, so as to eliminate the influence of data dimensions.

[0063] The risk probability unit is used to construct a gas leak location model based on a neural network through graph convolutional layers, LSTM layers, and fully connected network layers, respectively. The dynamic safety threshold of methane concentration and the concentration change rate threshold output by the risk assessment module are used as input constraints of the model, and the output is a three-dimensional vector of interface number, risk probability, and leakage rate.

[0064] This can be explained by the fact that gas pipeline leaks are prone to occur at various interfaces. By deploying point-type laser methane sensors at each interface, the risk of gas leaks at each interface can be effectively collected. Therefore, this solution uses a risk assessment module to analyze the changes in methane concentration and methane concentration change rate at each gas pipeline interface to assess and determine gas leaks at each interface. Secondly, using a gas leak location module, based on the assessment of the risk assessment module, the gas leak is located and deeply verified to avoid false alarms caused by gas diffusion at non-leakage nodes and reduce "warning fatigue" caused by frequent invalid warnings. On this basis, the valve self-adjustment module dynamically adjusts the valve opening degree at both ends of the interface of the suspected leak point according to the judgment data, which can not only reduce the risk of leakage, but also assist in accurate location through parameter changes during the adjustment process. Finally, by combining historical judgment data, on-site measurement results, and F1-Score stability trend curves, experts are assisted in evaluating the performance of the location model to achieve closed-loop optimization of the system, thereby comprehensively ensuring the intelligent control capability and operational reliability of the gas smart valve for gas transmission.

[0065] The method of installing three point-type laser methane sensors at a 120-degree angle in a ring at each interface of the gas transmission pipeline to collect methane concentration values ​​at different locations of the interface specifically includes:

[0066] Three point-type laser methane sensors are installed at each interface of the gas transmission pipeline, arranged in a ring. The angle between adjacent sensors is 120 degrees, forming a cone-shaped monitoring area around the interface. This area is used to collect methane concentration values ​​at different locations at the interface. The axis of each point-type laser methane sensor forms a 45° angle with the pipeline axis.

[0067] Based on the Kalman filter algorithm, the collected methane concentration is filtered and denoised to eliminate the influence of outlier peaks.

[0068] This can be explained by the fact that during gas transmission through pipelines, gas diffusion at leak points is affected by axial diffusion, radial injection, and eddy current diffusion. Single-point laser methane sensor monitoring may lead to incomplete data and false alarms. Therefore, this solution installs three point-type laser methane sensors at each interface of the gas transmission pipeline, arranged in a ring with a 120-degree angle between adjacent sensors, forming a conical monitoring area around the interface. This ensures that when the gas flow diffuses axially along the pipeline, at least one sensor is on the main path of the gas flow, avoiding missed detections caused by unidirectional gas flow. Furthermore, each sensor... A radial area at the pipeline interface is designed to ensure that at least one sensor is within the direct monitoring range of the jet gas flow, regardless of the direction of the injection. This addresses the issue of single-point sensors potentially deviating from the injection direction. Furthermore, the three-point ring layout utilizes spatial redundancy to cover the random distribution characteristics of vortices, reducing monitoring blind spots caused by vortices being obstructed by mechanical structures. This improves the accuracy and reliability of methane concentration monitoring at the gas pipeline interface. The concentration data from the three sensors can form a spatial comparison (e.g., the sensor concentration at the vortex accumulation point is significantly higher than the other two points), helping the risk assessment module distinguish between normal diffusion and localized high concentrations caused by vortices, avoiding misjudgments of multi-point leaks.

[0069] The installation of data acquisition devices both inside and outside each interface of the gas transmission pipeline to obtain pipeline and environmental parameter data at the interfaces specifically includes:

[0070] The temperature, humidity, dust concentration, air velocity, and distance between adjacent interfaces of the gas transmission pipeline are collected to gather its external environmental parameters.

[0071] The pressure and velocity of the gas at each interface of the gas transmission pipeline, as well as the inner diameter, cross-sectional area and valve opening of the pipeline, are collected to obtain its internal pipeline parameters.

[0072] Based on the Kalman filter algorithm, the collected pipeline parameter and environmental parameter data are filtered and denoised to remove the influence of outlier peaks.

[0073] Based on the min-max normalization formula, the collected pipeline parameters and environmental parameters are normalized to eliminate the influence of data dimensions.

[0074] Hardware synchronization and clock alignment are employed to ensure consistent acquisition frequencies and data bit synchronization for methane concentration, pipeline, and environmental parameters, thereby guaranteeing data consistency.

[0075] It can be explained that when using data to assess and locate gas leak risks, it is necessary to ensure the reliability and temporal consistency of the data. Therefore, this solution uses the Kalman filter algorithm and the min-max normalization formula to preprocess the data to eliminate the influence of data peaks and dimensions (which can be regarded as the subsequent parameters being calculated as dimensionless numerical values). In this process, external environmental parameters and internal pipeline parameters are collected simultaneously. This captures the impact of the environment on gas diffusion while also covering the pipeline's own operating status, providing a complete data foundation for the risk assessment module to calculate the "dynamic safety threshold" and the location module to construct "node attributes".

[0076] Reference Figure 2 As shown, the assessment of whether there is a gas leak at each interface based on the real-time methane concentration values ​​collected by three point-type laser methane sensors and the calculated methane concentration change rate specifically includes:

[0077] Maintain a circular buffer of length 'a' for each sensor to store the methane concentration values ​​collected most recently ('a'). Each time new data arrives, discard the oldest value and store the latest value.

[0078] Methane concentration data were acquired simultaneously from three point-type laser methane sensors, and the absolute rate of change of methane concentration in the two most recent acquisitions was calculated.

[0079] Based on the statistics of historical absolute change rates of methane concentration, and using the mean and standard deviation of the change rate over the past 24 hours, a threshold for the absolute change rate of methane concentration is set.

[0080] Based on the methane concentration value and the rate of change of methane concentration, a risk assessment was conducted at each interface of the gas transmission pipeline using a primary screening and secondary verification approach.

[0081] Based on the results of the initial screening and secondary verification, the risk warning is divided into Level 1 warning, Level 2 warning, and warning suppression.

[0082] The initial screening specifically includes:

[0083] Based on historical monitoring data of methane concentration, based on 3 The criteria are used to calculate the dynamic safety threshold for methane concentration in the current time series.

[0084] Determine whether the methane concentration monitoring value of any of the three sensors exceeds the corresponding dynamic threshold. If it does, mark the interface as a suspected leak point; otherwise, determine that there is no risk.

[0085] Calculate the methane concentration difference between any two sensors. If at least two sensors exceed the threshold and the concentration difference between any two sensors is greater than half of the safety threshold of either of the two sensors, then proceed to the second-level verification for further judgment.

[0086] The secondary verification specifically includes:

[0087] Based on the absolute rate of change of methane concentration from the point laser methane sensor, determine whether there are at least two sensors whose current rate of change exceeds the rate of change threshold. If so, calculate the trend of methane concentration change monitored by multiple sensors. If not, make a judgment based on boundary conditions.

[0088] The specific boundary condition restrictions include:

[0089] If only one sensor's absolute rate of change in methane concentration exceeds the rate of change threshold, but other sensors do not exceed the rate of change threshold, a sensor self-test prompt will be issued, and the self-test function of that sensor will be triggered.

[0090] If the methane concentrations detected by the three sensors continue to rise but the absolute rate of change is below the threshold, a long-term trend analysis prompt will be made, and the analysis results will be updated at fixed time intervals.

[0091] The expression for the dynamic safety threshold of methane concentration under the current time series is:

[0092]

[0093] In the formula, For the first The point-source laser methane sensor in the first... The dynamic safety threshold of methane concentration at any given time. For the first Initial methane concentration threshold for a point-type laser methane sensor No. The concentration standard deviation of a point-type laser methane sensor. , These are the correction coefficients fitted using historical data. This is a coupling correction term for the influence of internal and external environmental parameters at each interface of the gas transmission pipeline, where... This is a collection of environmental parameters at each interface of the gas transmission pipeline. This is a collection of external environmental parameters at each interface of the gas transmission pipeline.

[0094] The risk warning system, which categorizes risk warnings into Level 1, Level 2, and Warning Suppression based on methane concentration values ​​and changes in methane concentration, specifically includes:

[0095] When the secondary verification result indicates a suspected leak, a primary warning is triggered. Combined with the subsequent judgment result from the gas leak location module, the valve is immediately shut off or ventilation is activated.

[0096] When the secondary verification result indicates a slow leak, a secondary warning is triggered. Combined with the subsequent judgment result from the gas leak location module, a potential leak risk warning and location information are issued.

[0097] When the secondary verification result indicates environmental interference or sensor malfunction, an early warning suppression is triggered, and the judgment is delayed for 10 seconds to avoid false alarms.

[0098] The explanation is that gas contains methane. After a gas leak, the methane concentration can be monitored using a point-type laser methane sensor, and the risk can be assessed based on the methane concentration value. However, due to the complex environment of gas pipeline transmission, simply monitoring methane concentration is insufficient for accurate risk assessment and can easily lead to false alarms. Therefore, this solution, based on methane concentration risk assessment, combines methane concentration change rate with primary screening and secondary verification to conduct risk assessments at each interface of the gas transmission pipeline. This improves the accuracy and reliability of gas leak risk assessment. Simultaneously, a risk warning response mechanism is established to effectively improve the reliability of risk warnings and avoid interference from invalid warnings.

[0099] The expression for the absolute change rate of methane concentration in the two most recent measurements is as follows:

[0100]

[0101] In the formula, For the first The point-source laser methane sensor in the first... The absolute rate of change of methane concentration at time t. For the first The point-source laser methane sensor in the first... methane concentration monitoring value at time, For the first The point-source laser methane sensor in the first... methane concentration monitoring value at time, The time interval between two samplings;

[0102] The expression for the threshold of absolute change rate of methane concentration is:

[0103]

[0104] In the formula, For the first The threshold for the absolute rate of change of methane concentration in a point-type laser methane sensor. The balance coefficients are determined by fitting historical leakage events using the least squares method. For the first The average rate of change of a point-type laser methane sensor over the past 24 hours. For the first Standard deviation of the rate of change of a point-type laser methane sensor over the past 24 hours. This is the standard deviation adjustment factor, which is adjusted according to the pipeline pressure intensity.

[0105] In gas transmission pipeline leak detection, the dynamic safety threshold for methane concentration adopts dynamic compensation with environmental coupling terms to cope with complex nonlinear environments, thereby improving the reliability of detection under complex operating conditions and reducing false alarm rate. The rate of change threshold uses a simplified statistical model to quickly respond to sudden fluctuations, focusing on short-term statistics and improving the efficiency of fluctuation analysis.

[0106] The expression for the changing trend of methane concentration monitored by the multi-sensor system is as follows:

[0107]

[0108] In the formula, To monitor the changing trend of methane concentration using multiple sensors, For the number of sensors, For the first The weights of the point-type laser methane sensors, where... ,like If so, it is judged as a suspected leak. If it is, it is determined to be a slow leak. If so, it is determined to be environmental interference. If so, it is determined that the sensor is malfunctioning;

[0109] Among them, the coupling correction term for the influence of internal and external environmental parameters at each interface of the gas transmission pipeline. It dynamically adjusts the threshold to adapt to complex working conditions by linearly weighting the pipeline parameters and external environmental parameters. The weighting coefficients are obtained by fitting the past 100 leakage events using the least squares method.

[0110] Reference Figure 3 As shown, the step of constructing a gas leak location model based on a neural network based on the data collected by the data acquisition module, and obtaining the probability of gas leak risk at each interface, specifically includes:

[0111] Based on the collected internal pipeline parameters and external environmental parameters, the node attributes of the gas pipeline map are constructed.

[0112] The node attributes of the gas pipeline map include: external environmental parameters such as temperature and humidity, dust concentration, and air velocity, as well as internal environmental parameters such as gas pressure, velocity, pipeline inner diameter, cross-sectional area, and valve opening.

[0113] The edge attributes of the gas pipeline map are constructed based on the straight-line distance between interfaces, the elevation difference between interfaces, the pipeline connection angle, and the flow direction indicator.

[0114] A gas pipeline graph is constructed based on the attributes of each interface node of the gas pipeline and the attributes of the edges with adjacent interfaces.

[0115] The parameter data, methane concentration, and methane concentration change rate in the gas pipeline map are normalized based on the min-max normalization formula to eliminate the influence of data dimensions.

[0116] A gas leak location model based on a neural network is constructed by using graph convolutional layers, LSTM layers, and fully connected network layers respectively. The dynamic safety threshold of methane concentration and the concentration change rate threshold output by the risk assessment module are used as input constraints of the model, and the output is a three-dimensional vector of interface number, risk probability, and leakage rate.

[0117] The neural network-based gas leak location model specifically includes:

[0118] Graph convolution satisfies , In the formula, For nodes The output feature vector of the graph convolutional layer It is the ReLU activation function. For nodes The set of neighboring nodes, For nodes The degree, For nodes The degree, For learnable weight matrix, For nodes The original attribute vector, For connecting nodes and The edge attribute vector, To incorporate node attribute features and edge attribute features Concatenate by dimension, where the node degree refers to the number of edges directly connected to that node;

[0119] The ReLU activation function is used to perform a nonlinear transformation on the concatenated feature vectors to output fused features, thereby enhancing the model's ability to distinguish leaked features.

[0120] LSTM layer satisfies , In the formula, For nodes The output temporal feature vector of the LSTM layer The time step of the LSTM layer refers to the time interval between each state update. For the risk assessment module, the nodes The output graded early warning results are encoded and converted into time-series label values;

[0121] The LSTM layer utilizes the spatial features of graph convolution output, the time steps of synchronous sensor sampling, and the graded early warning results (Level 1 warning / Level 2 warning / early warning suppression) output by the risk assessment module. It integrates the spatiotemporal features of nodes with risk level labels to identify the dynamic changes of gas leaks such as sudden pressure drops, gradual increases in concentration, or changes in other pipeline and environmental parameters, in order to distinguish between emergency leaks and slow leaks.

[0122] Fully connected network layer satisfies , In the formula, For nodes The corresponding three-dimensional position coordinates, after mapping, can be converted into actual pipe interface numbers. For nodes The probability of leakage for the corresponding interface number. For nodes The leakage rate of the corresponding interface number;

[0123] in, For nodes The corresponding three-dimensional position coordinates are used to calculate their Euclidean distance using a predefined pipe interface coordinate database. The interface with the smallest distance is the actual pipe interface number after mapping. For nodes The leakage risk probability of the corresponding interface number is directly calculated by the sigmoid activation function of the MLP output layer to obtain the leakage risk probability value (the value range is [0,1]). The closer the probability is to 1, the higher the leakage risk.

[0124] The loss function satisfies , In the formula, , , These are the position loss weighting coefficient, the velocity loss weighting coefficient, and the regularization strength coefficient, respectively. The L2 norm symbol, To predict the interface number of the leak point for a neural network-based gas leak location model. The actual interface number of the leak point. To predict gas leak rates using a neural network-based gas leak location model, To represent the actual rate of gas leakage, This is the set of all learnable parameters for a neural network-based gas leak location model.

[0125] This study focuses on the accuracy of leak location and the severity of leaks corresponding to different leak rates. It analyzes the differences between predicted and actual leak locations, the difference between predicted and actual leak rates, and regularizes all learnable parameters. The study also addresses outliers and local noise in the data through regularization, thereby comprehensively improving the accuracy and reliability of the neural network-based gas leak location model. Specifically, the parameters... , , The optimal value is determined by cross-validation using a validation set.

[0126] This can be explained by constructing node attributes using internal pipeline parameters and external environmental parameters, and edge attributes using straight-line distance between interfaces, elevation difference between interfaces, pipeline connection angles, and flow direction identifiers. This transforms the physical pipeline network into computer-recognizable graph structure data. By mapping the correlation between node and edge attributes through graph convolutional layers using the gas pipeline map, the accuracy and comprehensiveness of gas leak location can be effectively improved. Specifically, by encoding the graded early warning results (Level 1 warning / Level 2 warning / suppression warning) output by the risk assessment module as temporal labels, which are used as supplementary inputs to the temporal features of the LSTM layer, the location model can simultaneously incorporate the prior label of risk level (e.g., Level 1 warning corresponds to strong temporal features of "emergency leak", and Level 2 warning corresponds to weak temporal features of "slow leak") when learning spatiotemporal features, thus strengthening the coupling between gas leak risk judgment and location.

[0127] The process of prioritizing the adjustment of valve openings at both ends of high-probability interfaces based on the probability of gas leak risk, and combining feedback results from a neural network-based gas leak location model to confirm whether a leak point is identified, specifically includes:

[0128] The interfaces are sorted from high to low leakage probability. The valves at both ends of the high-probability interfaces are finely adjusted at a fixed angle. The methane concentration changes of the interface and adjacent interfaces are collected based on a point-type laser methane sensor.

[0129] If the methane concentration decreases as the valves at both ends are adjusted, the interface is identified as a leak point, and the abnormal location is reported. If the methane concentration difference is within the specified range, the interface is identified as not a leak point, the valve opening is restored, and the above operation is repeated for the next highest probability interface until the leak point is located.

[0130] This can be explained by dynamically adjusting the valve opening of the high-probability leak interface and monitoring the methane concentration response to achieve rapid verification and precise location of the leak point. Specifically, when prioritizing fine-tuning the valves at both ends of the high-probability interface at a fixed angle, the valve fine-tuning angle can be adaptively adjusted based on the leak rate prediction value output by the gas leak location model based on a neural network. This can be explained, but is not limited to, that if the predicted leak rate is ≥0.5 m³ / h, the valve fine-tuning angle is adjusted to 15°, and if the predicted leak rate is <0.5 m³ / h, it is adjusted to 5°. Alternatively, the valve angle can be fine-tuned by combining the predicted leak rate and methane concentration using a PID algorithm, thereby improving the accuracy and efficiency of valve control.

[0131] The process of using historical data and field measurement results from a neural network-based gas leak location model to calculate the system's precision, recall, and F1-Score, and to provide feedback on the stability and accuracy of the neural network-based gas leak location model, specifically includes:

[0132] Simultaneously acquire monitoring data from a point-type laser methane sensor, judgment results from a neural network-based gas leak location model, and actual on-site monitoring results to form comprehensive historical data;

[0133] Based on comprehensive historical data, a historical data sample set is created, and the number of true positive samples, false positive samples, and false negative samples is determined by comparative analysis of the data.

[0134] The sample set must cover the typical and abnormal operating conditions of the system. The number of true positive samples refers to the number of samples that are actually positive and are predicted to be positive. The number of false positive samples refers to the number of samples that are actually negative but are predicted to be positive. The number of false negative samples refers to the number of samples that are actually positive but are predicted to be negative.

[0135] Precision is calculated based on the number of true positive samples and false positive samples, recall is calculated based on the number of true positive samples and false negative samples, and then the overall value is calculated using the F1-Score formula;

[0136] Based on the F1-Score value and the fluctuation range of its continuous calculation results, the stability and accuracy of the system risk verification and positioning are judged, and the performance status of the positioning model is fed back synchronously.

[0137] The F1-Score ranges from 0 to 1. The closer it is to 1, the better the overall performance of the system's risk verification and location. The smaller the fluctuation of the continuous calculation results, the more stable it is.

[0138] The performance status of the synchronous feedback positioning model refers to the comprehensive evaluation of the stability and accuracy of the positioning model based on the monitoring data of the point laser methane sensor, the judgment results of the positioning model, and the trend curve of the actual on-site monitoring results over the past 30 days, combined with the historical stability trend curve of F1-Score, and assisted by expert experience.

[0139] This can be explained as follows: To ensure the accuracy and reliability of the system's risk assessment and leak point location for gas pipeline leaks, this solution utilizes historical data and field measurement results from a neural network-based gas leak location model to calculate the system's precision, recall, and F1-Score. This quantifies the system's assessment accuracy and comprehensively verifies its stability and accuracy. Specifically, precision avoids misidentifying non-leak points as leak points, reducing unnecessary maintenance costs. Secondly, recall avoids missing actual leak points, ensuring gas system safety. Finally, the F1-Score comprehensively evaluates the system's ability to balance accurate identification and comprehensive coverage, providing a quantitative basis for system optimization.

[0140] In summary, the advantages of this invention are: it effectively diagnoses and locates the risk of gas pipeline leakage, and intelligently controls the valve opening.

[0141] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A smart gas valve with a point laser methane sensor, characterized in that, The device installation module is used to install three point laser methane sensors at an angle of 120 degrees at each interface of the gas transmission pipeline to collect methane concentration values in different directions of the interface. The data acquisition module is used to set acquisition devices inside and outside each interface of the gas transmission pipeline to obtain pipeline parameter and environmental parameter data inside and outside the interface. The risk assessment module is used to assess whether there is gas leakage at each interface according to the methane concentration values collected by the three point laser methane sensors in real time, and the methane concentration values and the calculated methane concentration change rate. The gas leakage positioning module is used to construct a gas leakage positioning model based on a neural network according to the data collected by the data acquisition module to obtain the probability of gas leakage risk at each interface. The valve self-adjustment module is used to adjust the opening degree of the valve at both ends of the interface with a high probability of gas leakage risk, and confirm whether it is a leakage point according to the feedback result of the gas leakage positioning model based on a neural network. The historical data analysis module is used to calculate the precision, recall rate and F1-Score value of the gas leakage positioning model based on a neural network by using the historical data and the on-site measurement results, and feed back the stability and accuracy of the gas leakage positioning model based on a neural network. The risk assessment module specifically includes: The buffer window unit is used to maintain a ring buffer with a length of a for each sensor to store the last a collected methane concentration values, discard the oldest value and store the latest value every time new data arrives. The change rate unit is used to obtain the methane monitoring concentration data under the same sequence of the three point laser methane sensors, and calculate the absolute change rate of the methane concentration collected in the last two times. The change rate threshold unit is used to set the absolute change rate threshold of the methane concentration based on the mean and standard deviation of the change rate in the past 24 hours according to the statistics of the historical absolute change rate of the methane concentration. The risk assessment unit is used to assess the risk of each interface of the gas transmission pipeline according to the methane concentration value and the methane concentration change rate by using the primary screening and secondary verification method. The hierarchical early warning unit is used to divide the risk warning into primary warning, secondary warning and warning suppression according to the determination results of the primary screening and secondary verification. When the secondary verification result determines that it is a suspected leakage, the primary warning is triggered, and the valve is closed or ventilation is triggered immediately according to the determination result of the subsequent gas leakage positioning module. When the secondary verification result determines that it is a slow leakage, the secondary warning is triggered, and potential leakage risk prompt and positioning information are issued according to the determination result of the subsequent gas leakage positioning module. When the secondary verification result determines that it is an environmental disturbance or a sensor anomaly, the warning suppression is triggered, and the judgment is delayed for 10 seconds to avoid false positives. ​ 2. A smart gas valve with a point type laser methane sensor according to claim 1, characterized in that, The three point type laser methane sensors are arranged at an angle of 120 degrees at each interface of the gas transmission pipeline to collect the methane concentration values in different directions of the interface, and the specific method comprises the following steps: Three point type laser methane sensors are arranged at each interface of the gas transmission pipeline to form a conical monitoring area around the interface, and the angle between adjacent sensors is 120 degrees, which is used to collect the methane concentration values in different directions of the interface, wherein the axis of each point type laser methane sensor is at an angle of 45 degrees with the axis of the pipeline; The collected methane concentration is filtered and denoised based on the Kalman filtering algorithm to eliminate the influence of abnormal peak values.

3. A smart gas valve with a point type laser methane sensor according to claim 2, characterized in that, The collecting device is arranged inside and outside each interface of the gas transmission pipeline to obtain the pipeline parameter and environmental parameter data of the inside and outside of the interface, and the specific method comprises the following steps: The temperature and humidity, dust concentration, air flow rate and distance between adjacent interfaces of the gas transmission pipeline at each interface are obtained to collect the external environmental parameters; The pressure, flow rate, inner diameter, cross-sectional area and valve opening degree of the gas at each interface of the gas transmission pipeline are obtained to collect the internal pipeline parameters; The collected pipeline parameter and environmental parameter data are filtered and denoised based on the Kalman filtering algorithm to eliminate the influence of abnormal peak values. The collected pipeline parameter and environmental parameter data are normalized based on the min-max normalization formula to eliminate the influence of data dimension; Hardware synchronization and clock alignment are adopted to make the collection frequency of the methane concentration, pipeline and environmental parameters consistent and the data bits synchronized to ensure data consistency.

4. The intelligent gas valve with a point type laser methane sensor according to claim 3, characterized in that, The primary screening specifically comprises the following steps: Based on historical monitoring data of the methane concentration, based on 3 criteria, a dynamic safety threshold of the methane concentration at the current time is calculated; It is judged whether any methane concentration monitoring value of the three sensors exceeds the corresponding dynamic threshold value, if yes, the interface is marked as a suspected leakage point, and if no, it is determined that there is no risk; The methane concentration difference between any two sensors is calculated, if at least two sensors exceed the threshold value and the concentration difference between any two sensors is greater than half of the safety threshold value of any one of the two sensors, the secondary verification is further judged; The secondary verification specifically comprises the following steps: According to the absolute change rate of the methane concentration of the point type laser methane sensor, it is judged whether the current change rate of at least two sensors exceeds the change rate threshold value, if yes, the change trend of the multi-sensor monitored methane concentration is calculated, and if no, the boundary condition is combined for judgment; The boundary condition limitation specifically comprises the following steps: If only the absolute change rate of the methane concentration of one sensor exceeds the change rate threshold value, but the absolute change rate of the other sensors does not exceed the change rate threshold value, a sensor self-checking prompt is made, and the self-checking function of the sensor is triggered; If the methane concentrations monitored by the three sensors are continuously rising but the absolute change rates are all lower than the threshold value, a long-term trend analysis warm prompt is made, and the analysis result prompt is updated at a fixed time interval; The expression of the dynamic safety threshold value of the methane concentration at the current time sequence is: In the formula, For the first The point-source laser methane sensor in the first... The dynamic safety threshold of methane concentration at any given time. For the first Initial methane concentration threshold for a point-type laser methane sensor No. The concentration standard deviation of a point-type laser methane sensor. , These are the correction coefficients fitted using historical data. This is a coupling correction term for the influence of internal and external environmental parameters at each interface of the gas transmission pipeline, where... This is a collection of environmental parameters at each interface of the gas transmission pipeline. This is a set of external environmental parameters at each interface of the gas transmission pipeline.

5. A smart gas valve with a point type laser methane sensor according to claim 4, characterized in that, The gas leakage positioning module specifically comprises: A node attribute unit is configured to construct the node attribute of the gas pipeline graph based on the collected internal pipeline parameters and external environmental parameters. The node attribute of the gas pipeline map includes temperature and humidity, dust concentration, air flow rate, and external environmental parameters, and internal environmental parameters of the gas, such as pressure, flow rate, pipeline inner diameter, cross-sectional area and valve opening degree; The edge attribute unit is configured to construct edge attributes of the gas pipeline map according to the straight-line distance between interfaces, the elevation difference between interfaces, the pipeline connection angle and the flow direction identifier; The pipeline map unit is configured to construct the gas pipeline map based on the node attributes of each interface of the gas pipeline and the edge attributes of adjacent interfaces; The data normalization unit is configured to normalize the parameter data, the methane concentration and the methane concentration change rate in the gas pipeline map based on a min-max normalization formula, so as to eliminate the influence of the data dimension; The risk probability unit is configured to construct a neural network-based gas leakage positioning model through a graph convolution layer, an LSTM layer and a fully connected network layer respectively, take the methane concentration dynamic safety threshold and the concentration change rate threshold output by the risk assessment module as model input constraint conditions, and output a three-dimensional vector of the interface number, the risk probability and the leakage rate.

6. A smart gas valve with a point type laser methane sensor according to claim 5, characterized in that, The neural network-based gas leakage positioning model specifically includes: Graph convolution satisfies , In the formula, For nodes The output feature vector of the graph convolutional layer It is the ReLU activation function. For nodes The set of neighboring nodes, For nodes The degree, For nodes The degree, For learnable weight matrix, For nodes The original attribute vector, For connecting nodes and edge attribute vector, To incorporate node attribute features and edge attribute features Concatenate by dimension, where the node degree refers to the number of edges directly connected to that node; The ReLU activation function is used to perform nonlinear transformation on the spliced feature vector, output fusion features, and enhance the model's ability to distinguish leakage features; The LSTM layer satisfies , , wherein, is a node The output time sequence feature vector of the LSTM layer, is a time step of the LSTM layer, indicating a time interval when the state is updated each time, is a value of a time sequence label obtained by encoding a graded early warning result output by the risk assessment module on the node . The LSTM layer uses the spatial features output by the graph convolution, the time steps sampled by the synchronous sensor, and the hierarchical early warning results output by the risk assessment module, which include first-level warning / second-level warning / warning suppression, to fuse the node spatiotemporal features and risk level labels, identify the dynamic change characteristics of the gas leakage caused by pressure drop, concentration rise or other pipeline and environmental parameter changes, and distinguish between emergency leakage and slow leakage. The fully connected network layer satisfies , , wherein, is a node corresponding to a three-dimensional position coordinate, which can be converted into an actual pipe interface number after mapping, is a node corresponding to a leak risk probability of the interface number, is a node corresponding to a leak rate of the interface number; wherein, is the node the corresponding three-dimensional position coordinates, calculate its Euclidean distance through the pre-defined pipeline interface coordinate database, the interface with the smallest distance is the mapped actual pipeline interface number, is the node the corresponding interface number of the leakage risk probability, which is directly calculated by the sigmoid activation function of the MLP output layer, so as to obtain the leakage risk probability value, the value range of the leakage risk probability value is [0, 1], and the closer the probability is to 1, the higher the leakage risk is. The loss function satisfies , , wherein , , are a position loss weight coefficient, a rate loss weight coefficient and a regularization intensity coefficient, respectively, is an L2 norm symbol, is an interface number of a leakage point predicted by the neural network-based gas leakage positioning model, is a real interface number of a leakage point, is a gas leakage rate predicted by the neural network-based gas leakage positioning model, is a real gas leakage rate of a leakage, is a set of all learnable parameters of the neural network-based gas leakage positioning model; Wherein, by analyzing the loss of the predicted position and the real position, the loss of the predicted gas leakage rate and the real leakage rate, and regularizing all learnable parameters, the accuracy of the leakage point positioning, the leakage severity corresponding to the leakage rate are focused on, and the regularization generalization processing is performed on the abnormal values or local noise in the data, thereby comprehensively improving the accuracy and reliability of the gas leakage positioning model prediction and judgment based on the neural network, wherein the parameters , , The optimal value is determined by cross-validation of the validation set.

7. A smart gas valve with a point type laser methane sensor according to claim 6, characterized in that, The method of preferentially adjusting the valve opening degree at both ends of the interface with a high probability of gas leakage according to the gas leakage risk probability, and confirming whether it is a leakage point based on the feedback result of the neural network-based gas leakage positioning model specifically includes: The interfaces are sorted in descending order of leakage probability, the valves at both ends of the interface with a high probability of gas leakage are preferentially adjusted by a fixed angle, and the methane concentration changes of the interface and adjacent interfaces are collected based on the point laser methane sensor; If the methane concentration decreases with the adjustment of the valves at both ends, it is determined that the interface is a leakage point, and the abnormal position is reported; if the methane concentration difference is within a specified range, it is determined that the interface is not a leakage point, and the valve opening degree is restored. The above operation is repeated for the interface with the second highest probability of gas leakage until the leakage point is located.

8. The intelligent gas valve with a point type laser methane sensor according to claim 7, characterized in that, The method of calculating the precision, recall and F1-Score values of the system by using the historical data and the field measurement results of the neural network-based gas leakage positioning model to feed back the stability and accuracy of the neural network-based gas leakage positioning model specifically includes: The monitoring data of the point laser methane sensor, the judgment result of the neural network-based gas leakage positioning model and the field actual monitoring result are obtained at the same time sequence to form comprehensive historical data; Based on the comprehensive historical data, a historical data sample set is made, and the number of true positive samples, false positive samples and false negative samples is determined through data comparison and analysis; Wherein, the sample set needs to cover typical and abnormal working conditions of the system, the number of true positive samples refers to the number of actual positive samples and predicted positive samples, the number of false positive samples refers to the number of actual negative samples but predicted positive samples, and the number of false negative samples refers to the number of actual positive samples but predicted negative samples; The precision is calculated based on the number of true positive samples and the number of false positive samples, the recall is calculated based on the number of true positive samples and the number of false negative samples, and the comprehensive value is calculated through the F1-Score formula; According to the value of F1-Score, combined with the fluctuation range of the continuous calculation result, the stability and accuracy of the system risk verification and positioning are judged, and the performance state of the positioning model is fed back synchronously; Wherein, the value range of F1-Score is between 0 and 1, the closer to 1 indicates that the comprehensive performance of the system risk verification and positioning is better, and the smaller the fluctuation range of the continuous calculation result indicates the more stable; The performance state of the synchronous feedback positioning model refers to the trend curve of the monitoring data of the point laser methane sensor, the judgment result of the positioning model and the actual monitoring result in the field in the past 30 days, combined with the historical stability trend curve of F1-Score, to assist the expert experience to comprehensively evaluate the stability and accuracy of the positioning model.

Images