A building gas leakage monitoring method and monitoring device based on pressure signals

Abstract

The application discloses a building gas leakage monitoring method and device based on pressure signals, comprising collecting pressure data of simple leakage, leakage during gas use and simple ignition instant, calculating linear slope of pressure data; introducing pressure data into a comprehensive model for training, and outputting recognition results; loading the comprehensive model in the main control unit of the monitoring device to judge pressure data of the pipeline, if the judgment result is leakage, triggering periodic gas supply interruption, and if the difference between pressure characteristics of this detection and a certain detection is less than or equal to the pressure set threshold or the difference between time characteristics is less than or equal to the time set threshold, determining as leakage, calculating the average value of leakage according to the average value of pressure characteristics and the average value of time characteristics of the two times, and determining the leakage grade. The application can accurately distinguish between pipeline leakage and ignition state through training of the machine learning model, avoid false judgment of leakage, accurately monitor gas leakage and improve gas use safety.

Description

Technical Field

[0001] This invention relates to a method for monitoring building gas leaks based on pressure signals, and also to a monitoring device using this method. Background Technology

[0002] With the profound transformation and continuous optimization of my country's energy structure, natural gas is playing an increasingly important role in energy consumption, with its consumption showing a significant growth trend. As the main fuel source for household gas appliances, it is widely used in the daily lives of millions of families. However, natural gas poses potential safety risks in daily use, especially the risk of leakage, making it a major public safety issue that cannot be ignored.

[0003] Gas pipeline system leaks have multiple causes, primarily including aging and failure of pipeline equipment beyond its service life; construction quality defects such as misaligned seals and loose welds caused by process or human factors during new construction or repair projects; pipe wall corrosion and thinning caused by soil, water vapor, or chemical erosion; and accidental damage to pipelines caused by third-party construction excavation or other mechanical forces. Once such leaks occur, their consequences often have severe chain reactions: firstly, the leaked natural gas directly pollutes the surrounding air environment; more seriously, flammable and explosive gases, when reaching a certain concentration and mixing with air, are highly likely to explode upon encountering an ignition source. Such destructive events not only pose a fatal threat to public safety in the affected area but also cause heavy casualties and incalculable losses of public and private property, disrupting social order and economic operations.

[0004] Currently, leak monitoring in the downstream of urban gas pipeline networks (usually referring to medium and low pressure pipeline systems, including residential courtyard pipeline networks and branch pipelines connecting to users) mainly relies on the traditional manual inspection mode.

[0005] Currently, the common method for preventing natural gas leaks in residential homes is to install household combustible gas detectors. However, this solution has the following drawbacks:

[0006] (1) The alarm has obvious blind spot limitations in spatial coverage. That is, the alarm installed only in specific locations such as the kitchen cannot effectively detect gas leaks in areas where no equipment is installed, such as corridors, bathrooms, balconies, and basements.

[0007] (2) The alarm itself is limited by the technical boundaries of its effective monitoring radius and sensitivity.

[0008] (3) The alarm itself has the potential risk of failure (such as sensor failure, circuit failure, battery depletion, environmental interference, etc.).

[0009] The combination of these factors leads to blind spots and delays in monitoring leakage risks in key areas of the user end, making it impossible to ensure that all potential risks can be effectively identified and promptly warned. Summary of the Invention

[0010] The first objective of this invention is to provide a building gas leak monitoring method based on pressure signals that can accurately monitor gas leaks and improve the safety of gas use.

[0011] The first objective of this invention is achieved through the following technical measures: a method for monitoring building gas leaks based on pressure signals, characterized by comprising the following steps:

[0012] S1. Collect pressure data of the main gas pipeline after the pressure regulating box in the simulation experiment within a set time before and after the occurrence of three working conditions: simple leakage, leakage during gas use, and simple ignition, and perform noise reduction processing on the pressure data.

[0013] S2. Calculate the linear slope of the pressure data for several time intervals within the set time range;

[0014] S3. Based on the features constructed by the linear slope of these pressure data, combined with synthetic minority oversampling technology, the denoised pressure data in step S1 is used as a training set and imported into the comprehensive model formed by combining the voting classifier, random forest model, XGBoost model and CatBoost model for classification training. The trained comprehensive model is then used to output the identification results of no leakage and gas use, simple leakage, simple ignition or leakage during gas use.

[0015] S4. The integrated model is mounted in the main control unit of the monitoring device, which is installed on the main gas pipeline after the pressure regulating box;

[0016] S5. The integrated model classifies the pipeline pressure data to determine whether there is no leakage or gas use, simple leakage, simple ignition or leakage during gas use. If the judgment result is leakage, proceed to step S6; otherwise, repeat this step.

[0017] S6. Trigger periodic gas supply interruption, and resume gas supply when the pressure is lower than the preset pressure required for gas use or when the preset time is reached. If the pressure is lower than the preset pressure, record the time characteristics and pressure characteristics; if the preset time is reached, record the pressure characteristics; this is one detection process.

[0018] S7. After each testing process, compare the time and pressure characteristics recorded in this testing process with the time and pressure characteristics recorded for all previous testing processes. If the difference in pressure characteristics is less than or equal to the pressure set threshold, or the difference in time characteristics is less than or equal to the time set threshold, then a leak is identified. Calculate the average leakage amount Q based on the average of the two pressure and time characteristics. avg Otherwise, it is judged as a suspected leak;

[0019] S8. Determine the leakage level:

[0020] No leakage: Q avg ≤0.02m 3 / h, leakage level is 0;

[0021] Minor leak: 0.02m 3 / h<Q avg ≤0.05m 3 / h, leakage level is 1;

[0022] Moderate leakage: 0.05m 3 / h<Q avg ≤0.1m 3 / h, leakage level 2;

[0023] Significant leak: 0.1m 3 / h <Q avg ≤0.5m 3 / h, leakage level 3;

[0024] Serious Leakage: Q avg >0.5m 3 / h, leakage level 4;

[0025] Suspected leak: Leakage level 5;

[0026] S9. Output the judgment result.

[0027] This invention, through training a machine learning model, can accurately distinguish between pipeline leaks and ignition states, avoiding incorrect leak detection. It can accurately monitor gas leaks, improve gas usage safety, and provide a basis for the maintenance and accident early warning of low-pressure gas pipeline networks in buildings. It is applicable to the field of safe operation of low-pressure gas pipelines. Moreover, this invention adopts a multi-cycle feature comparison method, repeatedly shutting off valves for short periods to record actual pressure and time characteristics. Based on feature similarity, it determines whether there is a leak in the building pipeline, and infers the average leakage flow rate during the shutdown period based on feature values. Therefore, this invention uses a dual active and passive detection algorithm to achieve comprehensive leak monitoring.

[0028] The pressure setting threshold of this invention is 20 Pa, and the time setting threshold is 5 seconds.

[0029] In step S7 of this invention, the average leakage amount Q is calculated. avg :

[0030] ;

[0031] In the formula, Q avg —Average volumetric flow rate of leakage, in cubic meters 3 / h; V1—pipe capacity, unit m 3 ; — Average pressure characteristic value, in Pa; —Time characteristic average value, in seconds; P0—Reference pressure, taken as 101.325 kPa.

[0032] In step S2 of this invention, the set time is 5 seconds, the set time range is 10 seconds, and the linear slope of the pressure data for several time intervals is adopted by the overall slope of the set time range representing the overall pressure situation within the set time range, the slope of the pressure drop segment representing the occurrence of a pressure drop, and the slope of the auxiliary judgment time interval representing the start of gas leakage or ignition. The linear slope of the pressure data for each time interval is fitted by the least squares method.

[0033] The second objective of this invention is to provide a monitoring device that uses the above-described pressure signal-based building gas leak monitoring method.

[0034] The second objective of this invention is achieved through the following technical measures: a monitoring device using the above-mentioned building gas leak monitoring method based on pressure signals, characterized in that it includes a housing, a display mounted on the housing, a connecting pipe disposed in the housing, a solenoid valve, a pressure sensor, a power module, and a main control unit equipped with a comprehensive model. The solenoid valve and the pressure sensor are disposed on the connecting pipe. The main control unit is connected to the pressure sensor, the solenoid valve, the display, and the power module respectively. The main control unit receives pressure data transmitted by the pressure sensor and performs calculations. The two ends of the connecting pipe form an air inlet and an air outlet on the housing. The air inlet and air outlet are connected to the main gas pipeline after the pressure regulating box.

[0035] The maximum working pressure of the pressure sensor described in this invention is 10 kPa.

[0036] The main control unit of this invention communicates with the server via a wireless communication module in order to upload the calculation results of the leakage level judgment to the server.

[0037] The power module of the present invention supports dual power supply modes of AC power and backup battery. It is powered by 220V AC power, and is powered by battery power after power failure and is powered by battery power after power is restored.

[0038] Compared with the prior art, the present invention has the following significant effects:

[0039] (1) By training a machine learning model, this invention can accurately distinguish between pipeline leakage and ignition status, avoid misjudging leakage, accurately monitor gas leakage, improve the safety of gas use, provide a basis for the maintenance and accident early warning of low-pressure gas pipeline networks in buildings, and is applicable to the field of safe operation of low-pressure gas pipelines. Moreover, by adopting a multi-cycle feature comparison method, the valve is shut off multiple times for short periods to record the actual pressure and time characteristics, and the presence of leakage in the building pipeline is determined based on feature similarity. The average leakage flow rate during the shut-off period is inferred based on the feature value. Therefore, this invention adopts a dual detection algorithm of active and passive detection to achieve all-round leakage monitoring.

[0040] (2) The monitoring device of this invention is installed on the main gas pipeline after the pressure regulating box. It can record the gas pressure in the pipeline in real time, and the gas supply can be cut off and restored by a solenoid valve. The pressure and detection information is uploaded by a wireless communication module. This invention detects pipeline pressure data through a pressure sensor, which can effectively identify the gas consumption and leakage of downstream pipelines, and can automatically monitor periodically to ensure the safe operation of the pipeline. The solenoid valve can control the pipeline to cut off the gas supply for leakage detection during maintenance. This invention has the characteristics of low cost, high reliability, and low false alarm rate, and can play a role in detecting leakage in the pipeline downstream of the pressure regulating box outlet.

[0041] (3) The monitoring device of the present invention can provide a basis for the maintenance and accident early warning of low-pressure gas pipeline network in buildings and is applicable to the field of safe operation of low-pressure gas pipelines.

[0042] (4) This invention innovatively integrates active disconnection and data upload functions to build an active security protection system.

[0043] (5) The overall solution of the monitoring device used in this invention has the advantages of low deployment cost, strong compatibility and easy maintenance, and is suitable for widespread promotion and application. Attached Figure Description

[0044] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0045] Figure 1 This is a schematic diagram illustrating the linear slope calculation of pressure data in this invention;

[0046] Figure 2 These are recall-precision curves for different categories of this invention;

[0047] Figure 3 This is a graph showing the true-false positive rate curves for different categories of positive cases in this invention;

[0048] Figure 4This is a schematic diagram of the confusion matrix between actual and predicted values ​​in this invention;

[0049] Figure 5 This is one of the structural diagrams of the monitoring device of the present invention;

[0050] Figure 6 This is the second structural diagram of the monitoring device of the present invention.

[0051] In the diagram: 1-Housing, 2-Display, 3-Connecting pipe, 4-Solenoid valve, 5-Pressure sensor, 6-Main control unit, 7-Inlet port, 8-Outlet port, 9-Power cord. Detailed Implementation

[0052] The present invention will be further described below through specific embodiments, but this is not a limitation of the present invention. Those skilled in the art can make various modifications or improvements based on the basic idea of ​​the present invention, but as long as they do not depart from the basic idea of ​​the present invention, they are all within the protection scope of the present invention.

[0053] like Figures 1-4 As shown, the present invention provides a method for monitoring building gas leaks based on pressure signals, comprising the following steps:

[0054] S1, such as Figure 1 As shown, pressure data of the main gas pipeline after the pressure regulating box was collected 5 seconds before and after the occurrence of three working conditions: simple leakage, leakage during gas use, and simple ignition in the simulation experiment (a total of 10 seconds). Pressure data was recorded every 0.1 seconds, and EMD signal denoising technology was used to denoise the negative pressure wave.

[0055] S2. Calculate the linear slope of the pressure data for several time intervals, i.e.:

[0056] K1: Total slope of the data from 0 to 10 seconds;

[0057] K2: Slope of the data from 0 to 6 seconds;

[0058] K3: Slope of the 6-10 second data;

[0059] K4: Slope of the data from 4.5 to 6.5 seconds;

[0060] K i , i=1,2,3,…,10: Pressure difference between second 0 and second i, where second 0 refers to the start time in the 10-second data.

[0061] All slopes are linear slopes of the pressure data for that time interval fitted using the least squares method. K1 represents the overall pressure situation during that time period. When K1 approaches 0, it indicates stable pressure. When K1 is negative and has a large absolute value, it indicates a significant pressure drop, which may indicate a leak or ignition. K4 represents the pressure drop segment, indicating a decrease in pressure. K2 and K3 assist K4 in determining whether a leak or ignition has begun. i It helps determine whether the problem is ignition or leakage.

[0062] The linear slopes of the pressure data for each time interval are shown in Table 1:

[0063] Table 1

[0064]

[0065] S3. Based on the linear slope of these pressure data, features are constructed (features and original pressure data are used to train the machine learning algorithm model; features are "another reflection of the original data" selected from a portion of the data by experience and machine learning itself; the following training involves inputting all the original data and these features into the algorithm, which calculates the features of these data in turn, and finally discusses the results based on the features). Combined with synthetic minority oversampling technology, the denoised pressure data from step S1 is used as the training set and imported into a comprehensive model composed of a voting classifier combining a random forest model, an XGBoost model, and a CatBoost model for classification training. The output is the identification results of no leakage and gas use (gas use includes ignition within a time period and continuous combustion after ignition before the start time point, labeled 0), simple leakage (label 1), simple ignition (label 2), or leakage during gas use (label 3); the classification performance evaluation results of the comprehensive model on the test set are shown in Table 2. The table details the model's identification capabilities for four different operating conditions: no leakage and normal gas use (label 0), simple leakage (label 1), simple ignition (label 2), and leakage during gas use (label 3). Evaluation metrics included accuracy, recall, and F1 score for each class. Data showed that the integrated model performed evenly across all classes, achieving a particularly high level of overall accuracy. This demonstrates that, combined with the voting classifier, the model effectively distinguishes between normal pipeline use, ignition fluctuations, and actual leaks, exhibiting good generalization ability and reliability.

[0066] Table 2

[0067]

[0068] a. Accuracy: The proportion of samples that are correctly predicted out of the total number of samples.

[0069] ;

[0070] b. Recall: The proportion of all true positive samples that are correctly predicted by the model.

[0071] ;

[0072] c. Precision: The proportion of true positives among all samples predicted as positive by the model.

[0073] ;

[0074] d. F1 score: The harmonic mean of precision and recall, used to measure the model’s overall performance in identifying and predicting positive classes.

[0075] ;

[0076] Table 3 defines the confusion matrix used to calculate the performance metrics of the model in this invention. This table shows four correspondences between the predicted and true classes: true positive (TP, where the predicted class is also positive), false negative (FN, where the predicted class is also positive), false positive (FP, where the predicted class is also negative), and true negative (TN, where the predicted class is also negative). Based on the fundamental parameters defined in Table 3, and combined with the specific calculation formulas for accuracy, recall, precision, and F1 score, the performance evaluation data in Table 2 is provided with a mathematical basis.

[0077] Table 3

[0078]

[0079] Synthetic minority oversampling is an existing technique that involves randomly selecting several samples from the nearest neighbor samples for each sample in a minority class, and then inserting new synthetic samples between these samples and the selected samples.

[0080] The integrated model of this invention combines three independent models using a voting classifier. If combining the predictions of three predictors (random forest, XGBoost, and CatBoost models) yields a better prediction than the best single predictor, this set of predictors is called an ensemble, the technique is called ensemble learning, and an ensemble learning algorithm is called an ensemble method (such as a voting classifier, where voting is based on the majority of the decisions from the internal logic outputs of the three models). In other words, ensemble learning does not rely on a single, potentially limited model, but rather combines the predictions of multiple predictors to achieve more accurate and stable predictions than any single component model.

[0081] Figure 4 This is a confusion matrix used to evaluate the performance of the classification model of this invention. The rows of the confusion matrix represent the true class (actual value) of the samples, the columns represent the predicted class (predicted value) of the model, and the values ​​on the diagonal represent the number of correctly classified samples. For example... Figure 4 As shown, the model accurately identifies "no leaks and gas usage" (label 0) (19,894 samples correctly classified), and also has good distinguishing ability between "simple leaks" (label 1) and "simple ignition" (label 2). However, for the critical safety risk scenario "leakage during gas usage" (label 3), the model exhibits some missed detections (misclassified as other categories) and false alarms (other categories misclassified as this type), indicating that identifying this complex state is the main challenge. This figure clearly demonstrates the overall effectiveness of the model and definite directions for optimization.

[0082] S4. The integrated model is mounted in the main control unit of the monitoring device, which is installed on the main gas pipeline after the pressure regulating box;

[0083] S5. The integrated model classifies the pipeline pressure data to determine whether it is leak-free and gas usage (label 0), simple leak (label 1), simple ignition (label 2), or leak during gas usage (label 3). If the judgment result is a leak, proceed to S6; otherwise, repeat this step.

[0084] S6. Trigger periodic gas supply interruptions and restore gas supply when the pressure is lower than the preset pressure required for gas use or when the preset time is reached (to ensure normal gas use, the pressure and time for gas supply restoration are preset). If the pressure is lower than the preset pressure, record the time and pressure characteristics; if the preset time is reached, record the pressure characteristics; this is one detection process.

[0085] S7. After each testing process, compare the time and pressure characteristics recorded in this testing process with the time and pressure characteristics of all previously recorded testing processes. If the difference in pressure characteristics is less than or equal to the pressure set threshold of 20 Pa or the difference in time characteristics is less than or equal to the time set threshold of 5 seconds, a leak is identified. Calculate the average leakage amount Q based on the average of the two pressure and time characteristics. avg Otherwise, it is judged as a suspected leak.

[0086] ;

[0087] In the formula, Q avg —Average leakage rate, in meters 3 / h; V1—pipe capacity, unit m 3 ; — Average pressure characteristic value, in Pa; —Time characteristic average value, in seconds; P0—Reference pressure, taken as 101.325 kPa.

[0088] S8. Determine the leakage level:

[0089] No leakage: Q avg ≤0.02m 3 / h, leakage level is 0;

[0090] Minor leak: 0.02m 3 / h<Q avg ≤0.05m 3 / h, leakage level is 1;

[0091] Moderate leakage: 0.05m 3 / h<Q avg ≤0.1m 3 / h, leakage level 2;

[0092] Significant leak: 0.1m 3 / h <Q avg ≤0.5m 3 / h, leakage level 3;

[0093] Serious Leakage: Q avg >0.5m 3 / h, leakage level 4;

[0094] Suspected leak: Leakage level 5;

[0095] S9. Output the judgment result.

[0096] Steps S6 to S9 of the present invention employ a multi-cycle feature comparison method, which records the actual pressure and time characteristics by repeatedly shutting off the valve for short periods of time. Based on the feature similarity, it is determined whether there is a leak in the building pipeline, and the average leakage flow rate during the shut-off period is inferred based on the feature value.

[0097] like Figure 5 , Figure 6 As shown, a monitoring device using the above-mentioned pressure signal-based building gas leak monitoring method includes a housing 1, a display 2 mounted on the housing 1, a connecting pipe 3 disposed in the housing 1, a solenoid valve 4, a pressure sensor 5, a power module, and a main control unit 6 equipped with a comprehensive model. The solenoid valve 4 and the pressure sensor 5 are disposed on the connecting pipe 3. The main control unit 6 is connected to the pressure sensor 5, the solenoid valve 4, the display 2, and the power module, respectively. The two ends of the connecting pipe 3 form an air inlet 7 and an air outlet 8 on the housing 1. The air inlet 7 and the air outlet 8 are connected to the main gas pipeline after the pressure regulating box.

[0098] Pressure sensor 5 has a range adapted to low-pressure pipeline conditions and is used to detect pipeline pressure data. Display 2 shows the current pipeline status and pressure fluctuations, and allows for viewing calculation parameters. Solenoid valve 4 is installed on the main gas pipeline after the pressure regulating box; its function is to cut off the main gas supply under specific circumstances. The valve is a specially customized model and is threaded to the pipeline. All the above instruments are explosion-proof.

[0099] The main control unit 6 can be powered by a 220V AC power supply and is connected to the power supply via power cord 9. The main control unit 6 receives pressure data transmitted by pressure sensor 5 and performs calculations, and communicates with the server via a wireless communication module to upload the calculation results for judging the leakage level to the server.

[0100] The power module of this invention supports dual power supply modes of AC power and backup battery. It uses 220V AC power and switches to battery power after power failure and charges the battery power after power is restored.

[0101] Example:

[0102] Pipeline pressure data were obtained through simulation experiments for simple gas use, gas use with leakage, and the pressure 5 seconds before and after simple ignition. Approximately 200 sets of data were measured for each condition. One set of data (totaling 10 seconds, some values ​​are shown in Table 4) is given below for each condition. Pressure data for the remaining time periods, i.e., data without gas use or leakage (label 0):

[0103] Table 4

[0104]

[0105] The collected stress data was imported into a combined model consisting of a random forest model, an XGBoost model, and a CatBoost model for classification training.

[0106] The training process of the model is optimized by setting the hyperparameters of the model through the following pre-pruning operations.

[0107] 1) Random Forest:

[0108] Number of trees in the forest: 180 trees; minimum number of samples required to split an internal node: 3; minimum number of samples required to split a leaf node: 3.

[0109] 2) XGBoost:

[0110] Number of trees: 180 trees, maximum tree depth: 50 layers, learning rate: 55%, subsampling ratio of training instances: 80%, feature subset ratio used when constructing trees: 80%.

[0111] 3) CatBoost:

[0112] Number of iterations: 800, learning rate: 50%.

[0113] With the above hyperparameter settings, the model can capture the features in the data well and has good general applicability. Moreover, the voting classifier integrates the three models into a comprehensive model.

[0114] The integrated model is mounted on Figure 5 and Figure 6 The main control unit chip of the monitoring device is connected to the main gas pipeline after the pressure regulating box to simulate leakage or ignition. The monitoring device can determine gas leakage in real time through a comprehensive model.

[0115] After a leak is detected, the leakage amount is estimated by shutting off the solenoid valve in the monitoring device multiple times (more than twice). After each shutdown, gas supply is restored when the gas pressure is lower than the preset pressure required for gas use (e.g., 1500 Pa) or when the preset time is reached (e.g., 60 s). If the pressure is lower than the preset pressure when gas supply is restored, the time characteristics (shutdown time) and pressure characteristics (pressure drop) are recorded; if the time reaches the preset time when gas supply is restored, the pressure characteristics (pressure drop) are recorded.

[0116] Compare the time and pressure characteristics with the previously recorded time and pressure characteristics respectively: if the difference in pressure characteristics is less than or equal to a set threshold (e.g., 20 Pa) or the difference in time characteristics is less than or equal to a set threshold (e.g., 5 s), then it is determined to be a leak.

[0117] Take a set of two data points (as shown in Table 5) when the pressure is lower than the preset pressure during gas supply restoration, and record the time characteristics (cut-off time) and pressure characteristics (pressure drop) for calculation:

[0118] Table 5

[0119]

[0120] The pressure at the end of both cut-offs was below 1500 Pa. The pressure drop and cut-off time were recorded, and the time difference was 29-24=5s, which determined to be a leak.

[0121] The average pressure drop from the first and second tests is ΔP, and the average cut-off time is Δt. The pipe volume V1 is 20L. Substituting these values ​​into the following formula:

[0122] ;

[0123] The average leakage rate was 0.035 m³. 3 / h, judged as a minor leak, leak level 1.

[0124] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the present invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A method for monitoring building gas leaks based on pressure signals, characterized in that... Includes the following steps: S1. Collect pressure data of the main gas pipeline after the pressure regulating box in the simulation experiment within a set time before and after the occurrence of three working conditions: simple leakage, leakage during gas use, and simple ignition, and perform noise reduction processing on the pressure data. S2. Calculate the linear slope of the pressure data for several time intervals within the set time range; the linear slope of the pressure data for several time intervals includes: the total slope of the set time range representing the overall pressure situation within the set time range, the slope of the pressure drop segment representing the pressure drop, and the slope of the auxiliary judgment time interval representing the start of gas leakage or ignition. The linear slope of the pressure data for each time interval is fitted using the least squares method. S3. Based on the features constructed by the linear slope of these pressure data, combined with synthetic minority oversampling technology, the denoised pressure data in step S1 is used as a training set and imported into the comprehensive model formed by combining the voting classifier, random forest model, XGBoost model and CatBoost model for classification training. The trained comprehensive model is then used to output the identification results of no leakage and gas use, simple leakage, simple ignition or leakage during gas use. S4. The integrated model is mounted in the main control unit of the monitoring device, which is installed on the main gas pipeline after the pressure regulating box; S5. The integrated model classifies the pipeline pressure data to determine whether there is no leakage or gas use, simple leakage, simple ignition or leakage during gas use. If the judgment result is leakage, proceed to step S6; otherwise, repeat this step. S6. Trigger periodic gas supply interruption, and resume gas supply when the pressure is lower than the preset pressure required for gas use or when the preset time is reached. If the pressure is lower than the preset pressure, record the time characteristics and pressure characteristics; if the preset time is reached, record the pressure characteristics; this is one detection process. S7. After each testing process, compare the time and pressure characteristics recorded in this testing process with the time and pressure characteristics of all previously recorded testing processes. If the difference in pressure characteristics is less than or equal to the pressure set threshold of 20 Pa or the difference in time characteristics is less than or equal to the time set threshold of 5 seconds, a leak is identified. Calculate the average leakage amount Q based on the average of the two pressure and time characteristics. avg : In the formula: Q avg —Average leakage rate, in meters 3 / h; V1—pipe capacity, unit m 3 ∆P—Pressure Characteristic average value, in Pa; ∆t—time characteristic average value, in s; P0—reference pressure, taken as 101.325 kPa; Otherwise, it is judged as a suspected leak; S8. Determine the leakage level: No leakage: Q avg ≤0.02m 3 / h, leakage level is 0; Minor leak: 0.02m 3 / h<Q avg ≤0.05m 3 / h, leakage level is 1; Moderate leakage: 0.05m 3 / h<Q avg ≤0.1m 3 / h, leakage level 2; Significant leak: 0.1m 3 / h <Q avg ≤0.5m 3 / h, leakage level 3; Serious Leakage: Q avg >0.5m 3 / h, leakage level 4; Suspected leak: Leakage level 5; S9. Output the judgment result.

2. The building gas leak monitoring method based on pressure signals according to claim 1, characterized in that: The set time is 5 seconds, and the set time range is 10 seconds.

Images