A method of gas leak source detection with additional confidence space fault tolerance

Abstract

The application discloses a gas leakage source tracing detection method with additional confidence space fault tolerance rate. On the basis of an atmospheric diffusion model, the application adds confidence space fault tolerance rate to the estimated positions of multiple gas leakage sources detected within a small interval time, combines a radio positioning method, establishes a fault tolerance space of a leakage point within continuous detection time, then carries out clustering to form a fault tolerance circle, carries out intersection processing on the fault tolerance circles formed at continuous detection time, obtains a fault tolerance space of an initial estimated leakage point, and finally carries out leakage point fault tolerance space filtering through environmental information to obtain a final leakage point estimated area. The reliability and scientificity of the leakage source positioning result are greatly improved.

Description

Technical Field

[0001] This invention relates to the field of gas leak source location technology, and specifically to a gas leak source tracing and detection method with additional confidence space tolerance. Background Technology

[0002] Rapid urbanization has driven a growing demand for clean air. Air pollution monitoring networks are being built across the country, with monitoring coverage gradually expanding. In improving air quality, accurately identifying pollution sources, timely monitoring changes in pollutant concentrations, and effectively issuing early warnings of air quality deterioration have become significant responsibilities. This is not only a prominent challenge in addressing environmental health risks but also crucial for achieving sustainable urban development.

[0003] Gas leak pollutants diffuse into the atmosphere, requiring atmospheric simulation to study their diffusion. Traditional atmospheric modeling methods include Lagrange models, computational fluid dynamics (CFD) models, and statistical models. Lagrange models, due to their unique particle characteristics, can effectively track turbulent eddies and large numbers of particles without distortion, making them helpful for studying concentration changes. While CFD models offer high accuracy, their complex parameters affect computational efficiency. Conversely, statistical models are relatively simple and computationally efficient. However, in complex scenarios, such as Gaussian plume and Gaussian dispersion models, they struggle to accurately describe atmospheric diffusion. Traditional static mechanism models present a trade-off between computational accuracy and efficiency, limiting their effectiveness in practical applications. To overcome this limitation, researchers have employed machine learning and data mining techniques to study gas diffusion. For example, Yeganeh et al. improved the traditional Support Vector Machine (SVM) and combined it with Partial Least Squares (PLS) to predict carbon monoxide concentrations in Tehran, achieving more accurate predictions and improving computational efficiency. In addition, Kopbayev A et al. highlighted the various applications of artificial neural networks in atmospheric science, including pattern classification and air quality prediction.

[0004] Existing gas leak source localization algorithms based on Gaussian plume atmospheric diffusion models focus on locating a single point, resulting in large errors, poor robustness, and a lack of regional applicability. Summary of the Invention

[0005] In view of this, the present invention provides a gas leak tracing and detection method with additional confidence space tolerance, which greatly improves the reliability and scientific validity of the leak source location results.

[0006] The gas leak tracing and detection method of the present invention with additional confidence space tolerance includes:

[0007] Step 1: Move the detection equipment to detect pollutants in the air; select N sets of continuous detection data, and calculate the suspected leakage sources for each set of detection data based on the gas diffusion model; where N≥3;

[0008] Step 2: For each group of detection data, the data is augmented based on the confidence interval of the detection equipment. Specifically, the factory confidence interval of each sensor of the detection equipment is used as the detection error range to generate detection data that includes random detection errors. The suspected leakage sources of the augmented data are obtained based on the gas diffusion model to realize the augmentation of the suspected leakage sources of the group.

[0009] Step 3: Cluster the expanded suspected leakage sources in each group to obtain N cluster circles; find the minimum circumcircle of all cluster circles.

[0010] Step 4: Divide the smallest circumcircle into 2m sector regions; where m is 3 to 8; count the number of suspected leakage sources in each sector region and sort them from largest to smallest; then, take the top m sector regions and make the following judgments:

[0011] If the first m sector regions are discontinuous, then the region containing the smallest circumcircle is considered to be a region where the leakage source is not located.

[0012] If the first m sector regions are continuous, then the first m sector regions are considered to be the areas where the leakage source is located.

[0013] Preferably, in step 1, the gas diffusion model is a Gaussian plume atmospheric diffusion model, a Lagrange model, or an Eulerian model.

[0014] Preferably, the gas diffusion model adopts a Gaussian plume atmospheric diffusion model; the confidence interval for methane concentration detection by the detection device is ±ΔC, and the confidence interval for wind speed detection is ±Δu; within these two confidence intervals, P methane concentration values ​​and P wind speed values ​​are randomly generated, denoted as C1, C2, ..., C P and u1, u2,...,u P By combining data in pairs according to the rules of permutation and combination, P×P sets of data are obtained. Based on the Gaussian plume atmospheric diffusion model, P×P suspected leakage sources are calculated respectively, thereby expanding the set of suspected leakage sources.

[0015] Preferably, in step 3, the clustering is performed using DBSCAN clustering, K-Means clustering, mean-shift clustering, or hierarchical clustering.

[0016] Preferably, in step 4, if the first m sector areas are continuous, the first m sector areas are filtered according to the ambient wind direction, and the filtered sector areas are the areas where the leakage source is located.

[0017] Ideally, the region where the leakage source is located is selected by using the smallest circumscribed circle diameter perpendicular to the wind direction as the boundary.

[0018] Beneficial effects:

[0019] (1) Based on the traditional Gaussian plume atmospheric diffusion model, this invention adds a confidence space tolerance rate to the estimated locations of multiple gas leak sources detected within a short interval, and combines it with radio positioning methods to propose a more accurate and scientific gas leak source tracing and detection model in urban environments. Based on the internal confidence interval of the gas detection equipment at the time of manufacture, the traditional Gaussian plume atmospheric diffusion model is used to establish a leak point tolerance space within continuous detection time. Then, clustering is performed to form tolerance circles. By intersecting the tolerance circles formed in continuous detection time, the initial estimated leak point tolerance space is obtained. Finally, environmental information is used to filter the estimated leak point tolerance space to obtain the final leak point prediction area. This greatly improves the reliability and scientific validity of the leak source location results.

[0020] (2) This invention is not limited to the Gaussian plume atmospheric diffusion model. Other gas diffusion models such as the Lagrange model and the Eulerian model can also be used to calculate the suspected leakage sources of each set of detection data.

[0021] (3) Further narrow down the range of leakage source location based on environmental factors. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the detection method of the present invention.

[0023] Figure 2 Flowchart for building a gas leak source tracing and detection model. Detailed Implementation

[0024] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0025] This invention, based on the traditional Gaussian plume atmospheric diffusion model, adds a confidence space tolerance rate to the estimated locations of multiple gas leak sources detected within a short time interval. Combined with radio positioning methods, it proposes a more accurate and scientific gas leak source tracing and detection method in urban environments. First, based on the internal confidence interval of the pollutant gas detection equipment at the time of manufacture, the invention uses the traditional Gaussian plume atmospheric diffusion model to establish a leak point tolerance space within continuous detection time. Then, clustering is performed to form tolerance circles. By intersecting the tolerance circles formed in continuous detection time, the initial estimated leak point tolerance space is obtained. Finally, environmental information is used to filter the estimated leak point tolerance space, resulting in the final leak point prediction area, greatly improving the reliability and scientific validity of the leak source location results.

[0026] The schematic diagram and flowchart of the detection method of this invention are as follows: Figure 1 and Figure 2 As shown, taking gas leak detection as an example, the specific steps are as follows:

[0027] The concentration of pollutant C(x,y,z) at any downwind point can be simulated using a Gaussian model of continuous point source diffusion in unbounded space.

[0028]

[0029] In this system, the leak point is taken as the origin, x is the downwind distance, y is the horizontal crosswind distance, z is the height of the gas sampling port of the ABB natural gas leak detection vehicle, set at 0.4m. Q is the methane emission intensity of the leak source, u is the wind speed, and σ is the methane emission intensity of the leak source. y ,σ z Let σ represent the Gaussian diffusion coefficient, which is the standard deviation of methane emission concentration from the natural gas leak source along the y-axis and z-axis, and follows a Gaussian distribution. This embodiment uses a commonly used empirical formula (Equation 2) to determine the Gaussian diffusion coefficient σ. x ,σ y ,σ z :

[0030]

[0031] Among them, a, b, c, and d are influencing factors, which are usually determined by atmospheric stability indices such as the Pasquill-Turner stability level, as shown in Table 1.

[0032] Table 1. Pasquale Atmospheric Stability Classification

[0033]

[0034]

[0035] The vehicle's hardware sensors continuously collect methane concentration C (mol / m³) at a rate of 2Hz. 3 ), wind speed u (m / s), wind direction θ (°). When the detection vehicle arrives at the suspected leak point area, three sets of continuous detection data are selected, and the coordinates of the three leak source rough location points A, B, and C are obtained using the Gaussian plume diffusion model (Equation 1), which are (x1,y1), (x2,y2), and (x3,y3), respectively.

[0036] Based on the internal confidence interval of the gas detection equipment at the time of manufacture, and using the traditional Gaussian plume atmospheric diffusion model, a fault tolerance space for the leak point is established within the continuous detection time.

[0037] Specifically, assuming the sensor's methane concentration detection error is ±ΔC and the wind speed detection error is ±Δu, within the detection errors of the two types of values, 10 sets of methane concentration values ​​and 10 sets of wind speed values ​​are randomly generated, denoted as C1, C2, ..., C10 and u1, u2,...,u 10 According to the rules of permutation and combination, 100 sets of data are obtained by pairwise combination. 100 possible leakage points are calculated by inversion according to formula (1). The above steps are performed on the three leakage sources A, B and C respectively to expand the group of suspected leakage sources.

[0038] Then, the three clusters of suspected leakage sources are clustered to form fault tolerance circles. Clustering algorithms such as DBSCAN clustering, K-Means clustering, mean shift clustering, and hierarchical clustering can be used for clustering.

[0039] This embodiment performs DBSCAN clustering on three clusters of data points. Taking 100 suspected leakage sources expanded from coarsely located point A as an example, the dataset is in the form of (X... a1 ,Y a1 ),(X a2 ,Y a2 ),...,(X a100 ,Y a100 The DBSCAN clustering algorithm steps are as follows:

[0040] (1) Distance metric

[0041] Using Euclidean distance as the metric:

[0042]

[0043] Where x i =(Xa i ,Ya i ) and x j =(Xa j ,Ya j () is a two-dimensional coordinate point.

[0044] (2) Neighborhood definition

[0045] For a point xi, its ∈ neighborhood is defined as

[0046] N ∈ (x i )={x j ∈X∣d(x i ,x j )≤∈} (4)

[0047] (3) Core point determination

[0048]

[0049] (4) Cluster expansion

[0050] From a core point x iInitially, add all points within the neighborhood of a new point to the cluster, and recursively check the points within the neighborhood of these new points:

[0051] C={x j ∈X∣x j For x i Density can reach (6)

[0052] Density reachability is defined as: the existence of a series of points x i ,x i+1 ,...,x j This makes each point a core point, and

[0053] (5) Calculation of the center (centroid) of the cluster circle

[0054] The center of a cluster can be obtained by calculating the average value of all points in the cluster. Assume the point set in cluster C is {(Xa1,Ya1),(Xa2,Ya2),...,(Xa...Ya2)}. k ,Ya k )}, center x a The calculation formula is:

[0055]

[0056] Where k is the number of points in the cluster.

[0057] (6) Calculation of the radius of the cluster circle

[0058] The radius of a cluster can be obtained by calculating the maximum distance from all points in the cluster to the center of the circle. The formula for calculating the radius 'a' is:

[0059]

[0060] Where μ=(μ x ,μ y () are the coordinates of the center of the circle.

[0061] Perform the above steps on the roughly located points B and C respectively, and obtain three circles with radii a, b, and c, and center coordinates (x, y, c). a ,y a ),(x b ,y b ),(x c ,y c Clustering circles A1, B1, C1

[0062] Then, find the minimum circumcircle D of the three clustering circles:

[0063] (1) Calculate the centroid G of the smallest circumcircle.

[0064]

[0065] (2) Calculate the distances from the centroid G to the centers of the three circles.

[0066]

[0067] (3) The side lengths of the triangles are:

[0068]

[0069] (4) Calculate the area of ​​the triangle

[0070]

[0071] (5) Calculate the radius R of the circumcircle of the triangle.

[0072]

[0073] (6) Calculate the center of the circumcircle of the triangle.

[0074] Using the geometric properties of the circumcircle of a triangle, the center of the circumcircle (x) d ,y d It can be calculated using the following formula:

[0075] First, calculate the coordinates (x, y) of the center of the circumcircle D. d ,y d ):

[0076]

[0077] Then determine the radius R of the smallest circumcircle. d

[0078]

[0079] The smallest circumcircle D is denoted as (x d ,y d ,R d This represents the initial estimated fault tolerance space for the leakage source.

[0080] Next, based on the principle of continuity of gas diffusion, the range of the leak source area is determined, specifically:

[0081] The initial estimated leakage source tolerance space (circumcircle D) obtained in the above steps is divided into 2m sector regions, where m is 3 to 8; in this embodiment, m = 6. The number of suspected leakage sources in each sector region is counted and sorted from most to least significant; the top six sector regions are selected for leakage source prediction.

[0082] If these six sector regions are discontinuous, then this possibility is ruled out, and the fault tolerance space (circumcircle D) is not the region where the leakage source is located.

[0083] If these six sector areas can be connected to form a semi-circular area, then the first m sector areas are the leakage source areas. The leakage source area can be further narrowed down based on environmental information (wind speed and direction) at the time of detection. Specifically, the sector areas located on the leeward side can be removed, using the smallest circumscribed circle diameter perpendicular to the wind direction as the boundary, and the sector areas located on the windward side of these six sector areas can be selected as the leakage source areas.

[0084] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A gas leak tracing and detection method with additional confidence space tolerance, characterized in that, include: Step 1: Move the detection equipment to detect pollutants in the air; Select N sets of continuous detection data and deduce the suspected leakage sources for each set of detection data based on the gas diffusion model; where N≥3; Step 2: For each group of detection data, the data is augmented based on the confidence interval of the detection equipment. Specifically, the factory confidence interval of each sensor of the detection equipment is used as the detection error range to generate detection data that includes random detection errors. The suspected leakage sources of the augmented data are obtained based on the gas diffusion model to realize the augmentation of the suspected leakage sources of the group. Step 3: Cluster the expanded suspected leakage sources in each group to obtain N cluster circles; find the minimum circumcircle of all cluster circles. Step 4: Divide the smallest circumcircle into 2m sector regions; where m is 3 to 8; count the number of suspected leakage sources in each sector region and sort them from largest to smallest; then, take the top m sector regions and make the following judgments: If the first m sector regions are discontinuous, then the region containing the smallest circumcircle is considered to be a region where the leakage source is not located. If the first m sector regions are continuous, then the first m sector regions are considered to be the areas where the leakage source is located.

2. The method as described in claim 1, characterized in that, In step 1, the gas diffusion model adopts a Gaussian plume atmospheric diffusion model, a Lagrange model, or an Eulerian model.

3. The method as described in claim 2, characterized in that, The gas diffusion model adopted is the Gaussian plume atmospheric diffusion model; the confidence interval for methane concentration detection by the detection equipment is ±ΔC, and the confidence interval for wind speed detection is ±Δu; within these two confidence intervals, P methane concentration values ​​and P wind speed values ​​are randomly generated, denoted as C1, C2, ..., C P and u1, u2,...,u P By combining data in pairs according to the rules of permutation and combination, P×P sets of data are obtained. Based on the Gaussian plume atmospheric diffusion model, P×P suspected leakage sources are calculated respectively, thereby expanding the set of suspected leakage sources.

4. The method as described in claim 1, characterized in that, In step 3, the clustering is performed using DBSCAN clustering, K-Means clustering, mean-shift clustering, or hierarchical clustering.

5. The method as described in claim 1, characterized in that, In step 4, if the first m sector areas are continuous, the first m sector areas are filtered according to the ambient wind direction, and the filtered sector areas are the areas where the leakage source is located.

6. The method as described in claim 5, characterized in that, Using the smallest circumscribed circle diameter perpendicular to the wind direction as the boundary, the fan-shaped area on the windward side is selected as the leakage source area.

Images