Method and system for laser detection of gas leaks for sealing environments

Abstract

This invention belongs to the field of safety measurement technology and discloses a laser detection method and system for gas leaks in sealed environments. The method includes: collecting operational data, environmental data, and monitoring data; using the operational data, environmental data, and a pre-set digital geographic grid of the monitoring area as input to a pre-trained gas migration model to obtain a dynamic background concentration field distribution map of the entire monitoring area; combining the spatial coordinates of each detection node to obtain the local background prediction value of each detection node; performing inversion analysis on the monitoring data to obtain the original gas volume fraction measurement value of each detection node, then subtracting the corresponding local background prediction value to obtain the compensated gas concentration residual value and analyzing it to obtain anomaly node clusters; and analyzing the anomaly node clusters with potential leak sources to obtain the location of the leak source. This invention enables the system to separate the weak residual signal characterizing the actual leak with high confidence from data submerged by strong background noise.

Description

Technical Field

[0001] This invention relates to the field of safety measurement technology, and more specifically, to a laser detection method and system for gas leaks in sealed environments. Background Technology

[0002] In core gas storage and transportation facilities such as gas storage depots and natural gas pipeline hubs, which contain multiple independent sealed units, gas leakage is a critical hidden danger that directly threatens operational safety and causes environmental risks and economic losses. Laser gas detection technology, especially tunable diode laser absorption spectroscopy, has been widely used in the field of gas leak monitoring due to its advantages such as high sensitivity, fast response, and non-contact measurement.

[0003] During the periodic injection and production of gas in a gas storage facility, pressure changes in the underground cavity cause trace amounts of gas to slowly diffuse to the surface through rock strata, wellbores, or sealed structures, resulting in global concentration fluctuations that vary with operating conditions. Existing monitoring systems typically treat the monitoring environment as a static or background noise-stable system, using static concentration baselines or fixed alarm thresholds. Under this mechanism, the system may misinterpret normal concentration fluctuations synchronized with the injection and production plan as diffuse leakage events, leading to frequent false alarms, disrupting operations, and wasting investigation resources. Conversely, it may also drown out genuine initial minute leak signals superimposed on a strong breathing background, resulting in missed detections and delaying accident response.

[0004] In view of this, the present invention proposes a laser detection method and system for gas leaks in sealed environments to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: a laser detection method for gas leakage in a sealed environment, comprising:

[0006] A sensor measurement network is formed by deploying detection nodes within the monitoring area to collect operational data, environmental data, and monitoring data from the detection nodes.

[0007] The operational data, environmental data, and a pre-set digital geographic grid of the monitoring area are used as inputs to a pre-trained gas migration model to obtain a dynamic background concentration field distribution map of the entire monitoring area, which serves as a benchmark for background concentration measurement at each spatial location.

[0008] The spatial coordinates of each detection node are matched and analyzed with the dynamic background concentration field distribution map to obtain the local background prediction value of each detection node.

[0009] Inversion analysis of the monitoring data is performed to obtain the original gas volume fraction measurement values ​​of each detection node; the corresponding local background prediction value is subtracted from the original gas volume fraction measurement values ​​of each detection node to obtain the gas concentration residual value after compensation.

[0010] By combining clustering algorithms to perform spatiotemporal analysis on the compensated gas concentration residual values ​​of all detection nodes in the entire monitoring area, abnormal node clusters are obtained. Spatial matching and correlation analysis are then performed on the abnormal node clusters and potential leakage sources to obtain the location of the leakage sources.

[0011] Furthermore, the operational data includes instantaneous injection / production flow rate, cavity pressure, and inventory change, wherein the inventory change is obtained by collecting the cumulative inventory before and after injection / production and calculating the difference between the cumulative inventory before and after injection / production; the environmental data includes wind speed and wind direction; and the monitoring data is the raw light intensity attenuation data of each detection node.

[0012] Furthermore, the methods for obtaining the original gas volume fraction measurements at each detection node include:

[0013] The measurement optical path is corrected based on the reference optical path digital voltage sequence in the original light intensity attenuation data of each detection node, and common noise is eliminated to obtain the corrected light intensity sequence.

[0014] The absorption rate as a function of time is calculated within each scanning cycle, and the optimal absorption spectral function is obtained by fitting the instantaneous wavelength of the laser.

[0015] The optimal absorption spectral function is integrated and its ratio to the product of the monitoring optical path length and the absorption line intensity is calculated to obtain the measured value of the original gas volume fraction at each detection node.

[0016] Furthermore, methods for obtaining the corrected light intensity sequence include:

[0017] Within the current scanning cycle of the laser, select a spectral range; calculate the average values ​​of the digital voltage sequence of the measurement optical path and the digital voltage sequence of the reference optical path within the spectral range, respectively.

[0018] The average values ​​of the digital voltage sequences of the measurement optical path and the reference optical path within the spectral range are subtracted from the digital voltage sequences of the measurement optical path and the reference optical path, respectively, to obtain the dominant sequence of the measured light intensity and the dominant sequence of the reference light intensity. At each sampling point, the ratio of the nominal value to the corresponding value of the dominant sequence of the reference light intensity is calculated as a correction factor. The product of each sequence value in the dominant sequence of the measured light intensity and the correction factor is calculated to obtain the corrected light intensity sequence.

[0019] Furthermore, methods for obtaining clusters of anomalous nodes include:

[0020] By combining clustering algorithms, spatial clustering is performed on the compensated gas concentration residual values ​​of all detection nodes in the entire monitoring area to obtain G spatial candidate clusters;

[0021] The detection node with the highest residual value of gas concentration after compensation in each spatial candidate cluster is taken as the central detection node, and the reverse trajectory simulation is performed in combination with environmental data;

[0022] Check whether the reverse trajectory region passes through or points to L potential leakage sources; L is the number of potential leakage sources; if the value of L is greater than 0, then take the spatial coordinates of each corresponding potential leakage source as the center, perform a forward simulation according to the Lagrange particle diffusion model, calculate the theoretical concentration distribution corresponding to the spatial coordinates of each detection node in the spatial candidate cluster when the potential leakage source produces a leak; calculate the Pearson correlation coefficient between the theoretical concentration distribution corresponding to the spatial coordinates of each detection node in the spatial candidate cluster and the actual observation residual value.

[0023] Spatial candidate clusters that simultaneously meet the anomaly determination conditions are identified as anomalous node clusters.

[0024] Furthermore, methods for obtaining spatial candidate clusters include:

[0025] For any detection node, calculate the mean and standard deviation of the compensated gas concentration residual value of the detection node within the sliding time window. If the mean is greater than three times the standard deviation, the corresponding detection node is judged to be abnormal within the current sliding time window. During the analysis period, if the number of consecutive sliding time windows in which a detection node is judged to be abnormal is greater than N or the cumulative proportion exceeds a preset ratio, the corresponding detection node is marked as a continuously abnormal detection node and recorded in the candidate detection node set. Based on the spatial coordinates of the detection node, a spatial clustering algorithm is used to cluster the detection nodes in the candidate detection node set to obtain G spatial candidate clusters, where k is a multiple of the standard deviation and N is the number of consecutive sliding time windows in which the detection node is judged to be abnormal.

[0026] Furthermore, methods for determining the location of the leak source include:

[0027] Obtain all abnormal node clusters, their corresponding detection node spatial coordinates, and the compensated gas concentration residual values;

[0028] Starting from the core node of the abnormal node cluster m, a reverse simulation is performed based on the Lagrange particle diffusion model to trace the origin trajectory of the air mass; potential leakage sources located in the upstream potential source region are extracted to form an initial candidate source set.

[0029] For each candidate source in the initial candidate source set, a forward simulation is performed to calculate the theoretical concentration value corresponding to each detection node in the anomaly node cluster m; the Pearson correlation coefficient between the theoretical concentration value corresponding to each detection node in the anomaly node cluster m and the actual observation residual value is calculated and normalized to obtain the spatial correlation score.

[0030] An exponential decay function is used to calculate the distance decay score based on the distance from the candidate source to the core node;

[0031] The spatial correlation score and distance decay score are weighted to obtain a comprehensive leakage score. The candidate source with the highest comprehensive leakage score is selected as the leakage source, and the corresponding spatial coordinates are output as the location of the leakage source.

[0032] Furthermore, in the preset digital geographic grid of the monitoring area, each grid point consists of spatial coordinates composed of the grid point's geographic coordinates and the corresponding elevation; the dynamic background concentration field distribution map provides the local background prediction value corresponding to each spatial coordinate point in grid form.

[0033] Furthermore, methods for obtaining the local background prediction value for each detection node include:

[0034] For each detection node, traverse each spatial coordinate point in the dynamic background concentration field distribution map. If the spatial coordinates of the detection node directly match the spatial coordinates of the point in the dynamic background concentration field distribution map, then use the local background prediction value corresponding to the spatial coordinates as the local background prediction value of the detection node; otherwise, obtain the spatial coordinates of the eight corner points of the grid cell to which the detection node belongs and the corresponding local background prediction values, and calculate the local background prediction value of the detection node through trilinear interpolation.

[0035] A laser detection system for gas leaks in sealed environments, comprising the implementation of the laser detection method for gas leaks in sealed environments, including:

[0036] Data acquisition module: A sensor measurement network is formed by detection nodes arranged in the monitoring area to collect the operating data, environmental data and monitoring data of the detection nodes;

[0037] The first analysis module takes the operational data, environmental data, and the pre-set digital geographic grid of the monitoring area as input to the pre-trained gas migration model to obtain the dynamic background concentration field distribution map of the entire monitoring area, which serves as the background concentration measurement benchmark at each spatial location.

[0038] The second analysis module matches and analyzes the spatial coordinates of each detection node with the dynamic background concentration field distribution map to obtain the local background prediction value of each detection node.

[0039] Concentration compensation module: Performs inversion analysis on monitoring data to obtain the original gas volume fraction measurement values ​​of each detection node; subtracts the corresponding local background prediction value from the original gas volume fraction measurement values ​​of each detection node to obtain the gas concentration residual value after compensation;

[0040] Leakage location module: Combines clustering algorithm to perform spatiotemporal analysis on the compensated gas concentration residual values ​​of all detection nodes in the entire monitoring area to obtain abnormal node clusters. Spatial matching and correlation analysis are performed on the abnormal node clusters with potential leak sources to obtain the location of the leak source.

[0041] The technical effects and advantages of the laser detection method and system for gas leaks in sealed environments according to this invention are as follows:

[0042] This invention proactively predicts and offsets the global, dynamic background concentration field caused by the normal operation of gas storage facilities through a digital model, thereby transforming the largest systemic interference in traditional monitoring into a known and removable benchmark signal. This enables the system to confidently separate the weak residual signal representing the actual leak from data submerged in strong background noise, solving the problem of false alarms and missed alarms caused by the inability of traditional static threshold methods to distinguish between operational emissions and accident leaks. Based on this, through spatiotemporal clustering of abnormal leak signals and reverse trajectory verification based on a physical diffusion model, it achieves accurate location and accountability for multiple leak sources, providing a dynamic, adaptive, and attributable new intelligent monitoring benchmark for the safe operation and risk management of gas storage facilities. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of the laser detection method for gas leakage in a sealed environment according to the present invention;

[0044] Figure 2 This is a schematic flowchart of the method for obtaining the original gas volume fraction measurement values ​​at each detection node according to the present invention;

[0045] Figure 3 This is a schematic diagram of the method for obtaining spatial candidate clusters according to the present invention;

[0046] Figure 4 This is a schematic diagram of the structure of the laser gas leak detection system for sealed environments according to the present invention. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] Example 1:

[0049] Please see Figure 1 As shown, this embodiment provides a laser detection method for gas leaks in a sealed environment, including:

[0050] A sensor measurement network is formed by detection nodes deployed within the monitoring area to collect operational data, environmental data, and monitoring data from the detection nodes. The operational data consists of real-time injection and production plans and gas storage condition data, including instantaneous injection and production flow rates, cavity pressure, and inventory changes. The inventory change is obtained by collecting the cumulative inventory before and after injection and calculating the difference between the cumulative inventory before and after injection. The environmental data consists of real-time environmental data obtained from meteorological stations, including wind speed and direction. The monitoring data consists of raw light intensity attenuation data collected from each detection node through an open laser gas monitoring network deployed within the monitoring area.

[0051] Operational and environmental data, along with a pre-defined digital geographic grid of the monitoring area, are used as inputs to a pre-trained gas migration model to obtain a dynamic background concentration field distribution map of the entire monitoring area. This map serves as the baseline for background concentration measurement at each spatial location. Within the pre-defined digital geographic grid of the monitoring area, each grid point consists of its geographic coordinates and corresponding elevation. The dynamic background concentration field distribution map provides the local background prediction value for each spatial coordinate point in a grid format. Operational data is the direct driving variable for quantifying the gas storage facility's respiration intensity and state. Combining operational data enables the system to predict, rather than only detect, background concentration changes caused by legitimate operations, for the first time, distinguishing dynamic interference. The foundation of the system; environmental data provides the dynamic conditions for the diffusion and transmission of the background concentration field on the surface, which is the key to the model's transformation of underground respiratory flux into a spatial concentration distribution map, enabling the predicted background field to have spatial authenticity; the pre-set digital geographic grid of the monitoring area provides a spatial carrier for the calculation and output of the dynamic background concentration field, elevating the model from point or path prediction to field prediction that can be spatially matched with the entire monitoring network, which is a prerequisite for achieving accurate spatial compensation; the dynamic background concentration field distribution map is a dynamic map used to offset systematic interference in real time. It represents the true background that is missing in traditional methods and changes with operating conditions, enabling subsequent comparison of monitoring data with it to directly separate abnormal leakage signals.

[0052] Training methods for gas transport models include:

[0053] Pre-collected B1 group transfer training data, which is precise data collected for model training under the condition that gas leakage has occurred in a known sealed environment; the transfer training data includes operational data and environmental data, as well as a pre-set digital geographic grid of the monitoring area and the actual background concentration field distribution map of the entire monitoring area. The actual background concentration field distribution map of the entire monitoring area can be generated by collecting full-area concentration data by gas sensors and then interpolating it with the digital geographic grid.

[0054] The gas migration model takes operational data, environmental data, and a pre-defined digital geographic grid of the monitoring area as input. The gas migration model outputs a dynamic background concentration field distribution map of the entire monitoring area. With the goal of minimizing the error between the output dynamic background concentration field distribution map and the actual background concentration field distribution map of the entire monitoring area, the backpropagation algorithm and gradient descent optimization algorithm are used to iteratively optimize the network parameters of the gas migration model until the gas migration model converges. The network parameters corresponding to the convergence of the gas migration model are obtained, and the gas migration model constructed with the corresponding network parameters is used as the trained gas migration model.

[0055] The spatial coordinates of each detection node are matched with the dynamic background concentration field distribution map to obtain the local background prediction value of each detection node.

[0056] Methods for obtaining the local background prediction value for each detection node include:

[0057] For each detection node, traverse each spatial coordinate point in the dynamic background concentration field distribution map. If the spatial coordinates of the detection node directly match the spatial coordinates of the point in the dynamic background concentration field distribution map, then use the local background prediction value corresponding to the spatial coordinate point as the local background prediction value of the detection node; otherwise, obtain the spatial coordinates of the eight corner points of the grid cell to which the detection node belongs and the corresponding local background prediction values, and calculate the local background prediction value of the detection node through trilinear interpolation; as detailed below:

[0058] Based on the minimum coordinate value among the eight corner points (i.e., the spatial coordinates of the lower left rear corner point) and the maximum coordinate value among the eight corner points (i.e., the spatial coordinates of the upper right front corner point), the spatial coordinates of the detection node are... Perform normalization calculations to obtain normalized coordinates. The local background prediction values ​​corresponding to the eight corner points are weighted and summed using normalized coordinates to calculate the local background prediction value corresponding to the detection node; for example, the formula for calculating the local background prediction value of the detection node is:

[0059] ;

[0060] in, The predicted local background value for the detection node; The local background prediction values ​​correspond to eight corner points, which are, in order, the lower left rear corner point, the lower right rear corner point, the lower left front corner point, the lower right front corner point, the upper left rear corner point, the upper right rear corner point, the upper left front corner point, and the upper right front corner point. The positive X-axis is from left to right, the positive Y-axis is from back to front, and the positive Z-axis is from bottom to top. Through spatial matching and interpolation, the continuous three-dimensional prediction field is accurately assigned to each discrete detection node, realizing the quantitative correlation from field to point, and laying the foundation for accurate compensation of each node.

[0061] Inverse analysis of the monitoring data yields the original gas volume fraction measurements at each detection node. Subtracting the corresponding local background prediction value from the original gas volume fraction measurements at each detection node provides the compensated gas concentration residual. By stripping away dynamic background concentration fluctuations, the residual value reflecting only abnormal leaks is obtained. This overcomes the limitation of traditional methods that treat the background as stationary noise, accurately focusing on leak signals and filtering out dynamic system interference caused by the operation of the monitored object.

[0062] Reference Figure 2 Methods for obtaining the original gas volume fraction measurements at each detection node include:

[0063] A known driving current, consisting of a high-frequency sinusoidal scanning signal and a low-frequency sawtooth scanning signal, is injected into the distributed feedback laser configured at each detection node. This current is used to enable the center wavelength of the laser output to continuously and periodically scan near a single absorption line of the target gas. The laser emits detection laser and reference laser.

[0064] The detection detector receives the detection laser output from the laser and collects the measurement voltage signal output by the detection detector. The laser emitted by the laser is divided into two paths: a measurement optical path and a reference optical path. The laser in the measurement optical path passes through the open atmospheric monitoring path and is received by the detection detector installed at the other end of the atmospheric monitoring path. The detection detector receives the light intensity signal and linearly converts it into a measurement voltage signal.

[0065] A reference detector receives a reference laser beam from a laser that passes through a pre-set sealed reference chamber and acquires the reference voltage signal output by the detector. The sealed reference chamber is a built-in, fixed-length chamber filled with a known low-concentration target gas or clean air. The laser beams from both the reference and measurement optical paths simultaneously pass through this chamber, which is filled with a reference gas of a known concentration (different from the gas stored in the gas reservoir). This reference gas is received by a reference detector at the other end of the chamber, which receives the light intensity signal and linearly converts it into a reference voltage signal. By synchronously monitoring a reference optical path unaffected by the ambient gas, a benchmark can be provided for identifying and eliminating common-mode noise from laser power fluctuations, temperature drift, and other device-related factors. This is the foundation for achieving high-precision single-point concentration measurement.

[0066] The measured voltage signal and the reference voltage signal are converted from analog to digital to obtain the digital voltage sequence of the measured optical path and the digital voltage sequence of the reference optical path. The obtained digital voltage sequence of the measured optical path and the digital voltage sequence of the reference optical path are the original light intensity attenuation data of each detection node obtained from the monitoring network.

[0067] The measurement optical path is corrected based on the reference optical path digital voltage sequence. Common noise caused by laser power fluctuations in the measurement optical path digital voltage sequence is eliminated to obtain the corrected optical intensity sequence.

[0068] Methods for obtaining the corrected light intensity sequence include:

[0069] Within the current scanning cycle of the laser, a spectral range without target gas absorption is selected. Specifically, two continuous spectral ranges are selected on either side of the characteristic absorption spectral line of the gas stored in the gas storage tank. The width of each spectral range can be set empirically. The spectral range must be completely outside the absorption spectral range of the target gas, and the distance between the edge of the spectral range and the center frequency of the absorption spectral line must be at least 2-3 times the full width at half maximum (FWHM) of the absorption line to ensure it is not affected by the slight influence of the absorption line's fins. The average values ​​of the digital voltage sequence of the measurement optical path and the digital voltage sequence of the reference optical path are calculated within the spectral range. This average value mainly includes fixed background factors such as detector dark current and circuit bias, isolating a clean instrument response range in the spectrum to subtract fixed background factors such as detector dark current and circuit offset. This ensures that the baseline for subsequent analysis is zeroed, improving the signal-to-noise ratio.

[0070] The measured light intensity dominant sequence and the reference light intensity dominant sequence are obtained by subtracting the average values ​​of the measured light path digital voltage sequence and the reference light path digital voltage sequence from the measured light path digital voltage sequence and the reference light path digital voltage sequence, respectively, within the spectral range.

[0071] At each sampling point, the ratio of the nominal value to the dominant reference light intensity sequence value corresponding to the sampling point is calculated as a correction factor; the nominal value is the average value of the dominant reference light intensity sequence value of the reference optical path throughout the entire scanning period.

[0072] The corrected intensity sequence is obtained by multiplying each value in the measured intensity dominant sequence by the correction factor. If the laser power drops instantaneously, both the measured and reference intensity dominant sequences will decrease proportionally. In this case, the correction factor will be greater than 1, proportionally amplifying the measured intensity dominant sequence to offset the power decrease. Ultimately, fluctuations in the corrected intensity sequence will primarily reflect changes caused by gas absorption in the measured optical path. By comparing the measured optical path signal with the reference optical path signal, the shared intensity changes experienced by both can be offset, ensuring the output signal reflects only the relative changes caused by gas absorption.

[0073] The absorption rate as a function of time is calculated for each scanning cycle, and the optimal absorption spectral function is obtained by fitting the instantaneous wavelength of the laser.

[0074] Methods for obtaining the optimal absorption spectral function include:

[0075] Within the same wavelength scanning period, a spectral range is selected, and the corrected light intensity sequence is polynomial-fitted within the spectral region to obtain the light intensity baseline sequence.

[0076] For each sampling point, the negative logarithm of the ratio of the corrected intensity sequence to the baseline intensity sequence is calculated to obtain the absorptivity. The absorptivity directly reflects the degree to which the laser is absorbed by gas molecules along the propagation path. The smaller the corrected intensity sequence, the greater the absorptivity.

[0077] The time corresponding to the sampling point index is mapped to the instantaneous wavelength value of the laser through the driving current.

[0078] Using each instantaneous laser wavelength as the independent variable and the corresponding absorption rate as the dependent variable, an absorption spectral function is constructed based on the Voigt linear function. The Voigt model is a well-known standard function characterizing the spectral line shape, so choosing the Voigt model can best fit the relationship between the absorption rate and each instantaneous laser wavelength.

[0079] The nonlinear least squares method is used to optimize the relevant parameters of the absorption spectral function with the goal of minimizing the sum of squared residuals between the function value of the absorption spectral function and the experimentally measured absorbance. Optionally, real-time temperature, air pressure and humidity data can be combined to physically correct the relevant parameters of the absorption spectral function, such as absorption line intensity and broadening, to further improve the accuracy of spectral fitting and thus obtain the optimal absorption spectral function.

[0080] The optimal absorption spectral function is integrated to obtain the integral area of ​​the absorption spectral line. The product of the path length of the monitoring optical path and the absorption line intensity is calculated, and then the ratio of the integral area of ​​the absorption spectral line to the product is calculated to obtain the original gas volume fraction measurement value of each detection node. Among them, the path length of the monitoring optical path is known and calibrated during installation, and the absorption line intensity is a physical constant related to the gas type, specific absorption spectral line and ambient temperature, which can be obtained by querying a standard spectral database.

[0081] By using dual-optical-path synchronous measurement to offset common noises such as laser power fluctuations, baseline extraction in the non-absorption region and Voigt line nonlinear fitting improve inversion accuracy and provide high-precision raw data for residual calculation; the complete light intensity, absorptivity and concentration link solves the problems of traditional inversion neglecting system noise and line matching.

[0082] By combining clustering algorithms to perform spatiotemporal analysis on the compensated gas concentration residual values ​​of all detection nodes in the entire monitoring area, abnormal node clusters are obtained. Spatial matching and correlation analysis are then performed on the abnormal node clusters and potential leakage sources to obtain the location of the leakage sources.

[0083] Methods for obtaining anomalous node clusters include:

[0084] For any detection node, calculate the mean and standard deviation of the compensated gas concentration residual values ​​of the detection node within the sliding time window. If the mean is greater than k times the standard deviation, the corresponding detection node is judged to be abnormal within the current sliding time window. During the analysis period, if the number of consecutive sliding time windows in which a detection node is judged to be abnormal is greater than N times or the cumulative proportion exceeds a preset ratio, the corresponding detection node is marked as a continuously abnormal detection node and added to the candidate detection node set. Here, k is a multiple of the standard deviation, usually taken as 2-3; N is the number of consecutive sliding time windows in which a detection node is judged to be abnormal, preferably 3-5. The preset ratio is obtained by statistically analyzing the distribution of the cumulative proportion of sliding time windows in which all detection nodes are judged to be abnormal during the analysis period in the initial calibration phase of the system or during long-term leak-free operation, and taking its statistical upper limit, such as the 95th percentile, as the preset ratio. In practical applications, the typical empirical value range of this preset ratio is 70% to 80%.

[0085] Obtain the spatial coordinates of each detection node in the candidate detection node set and the corresponding gas concentration residual value after compensation; based on the spatial coordinates of the detection nodes, use a spatial clustering algorithm to cluster the detection nodes in the candidate detection node set to obtain G spatial candidate clusters.

[0086] Reference Figure 3 Methods for obtaining spatial candidate clusters include:

[0087] Step 1: Traverse each detection node Pi in the candidate detection node set, and obtain other detection nodes whose Euclidean distance from detection node Pi is not greater than the neighborhood radius. If the total number of detection nodes contained in the neighborhood of detection node Pi is not less than the minimum number of points, then detection node Pi is marked as a core point; otherwise, it is marked as a boundary point. The neighborhood radius is determined according to the average deployment spacing of the monitoring network. The minimum number of points is the minimum number of detection nodes required to form a cluster, usually set to 2 or 3, to exclude isolated noise points.

[0088] Step 2: Starting from any unvisited detection node Pj, create a new cluster W and add Pj to cluster W. Traverse all detection nodes Ps in the neighborhood of Pj. If Ps has not been visited, mark Ps as visited. If Ps is a core node, add Ps and all unassigned detection nodes in the neighborhood of Ps to cluster W.

[0089] Step 3: Repeat step 2 until all detection nodes in the neighborhood of the current core point have been visited and allocated, at which point a complete cluster is generated.

[0090] Step 4: Select the next unvisited core point and repeat steps 2-3 until all core points have been visited.

[0091] Step 5: Remove all detection nodes that have not been assigned to any cluster to obtain G spatial candidate clusters.

[0092] The detection node with the highest residual value of gas concentration after compensation in each spatial candidate cluster is taken as the central detection node, and reverse trajectory simulation is performed in combination with environmental data.

[0093] Methods for simulating reverse trajectories include:

[0094] Set the simulation time step Δt1, the total simulation duration T, and the number of released virtual particles M. Δt1 is set according to wind speed and spatial resolution, usually 0.5-2 hours; T is set according to the gas diffusion time scale, usually 6-24 hours; M≥1000 to ensure the stability of trajectory statistics.

[0095] Starting from the spatial coordinates of the central detection node, and based on the collected real-time wind speed and direction data, driven by reverse wind direction and reverse wind speed, i.e., the wind direction is completely reversed, ,in, This refers to the reverse wind field velocity; The real-time wind field velocity is simulated in reverse according to the Lagrange particle diffusion model. Specifically, starting from the central detection node, M virtual particles are released, and the motion of each particle is described by the following equation: ;in, For particles in The position at that moment; Let be the position of the particle at time t; For particles in the reverse wind field The velocity at a given moment; The random turbulence term follows a normal distribution with a mean of 0 and a variance of 2K times Δt1, where K is the turbulence diffusion coefficient, which can be obtained from a table based on the atmospheric stability level. The motion paths of all particles during the simulation period are recorded, and the residence time or frequency of particles within each grid cell is statistically analyzed to generate a probability density distribution map, which represents the reverse trajectory region.

[0096] Check whether the reverse trajectory region passes through or points to L potential leakage sources; the specific steps are as follows:

[0097] If the probability density value of a grid cell containing a potential leak source exceeds a preset probability density threshold, the reverse trajectory region is considered to have passed through the corresponding potential leak source. The preset probability density threshold is set by analyzing the reverse trajectory simulation data during the past 30 days without leaks, obtaining the background mean and background standard deviation of the probability density of each grid cell, and then setting the threshold based on the background mean and background standard deviation.

[0098] L represents the number of potential leakage sources. If L is greater than 0, a forward simulation is performed based on the Lagrange particle diffusion model, centered on the spatial coordinates of each corresponding potential leakage source. The theoretical concentration distribution corresponding to the spatial coordinates of each detection node in the spatial candidate cluster is calculated when the potential leakage source leaks. The Pearson correlation coefficient between the theoretical concentration distribution corresponding to the spatial coordinates of each detection node in the spatial candidate cluster and the actual observation residual value is calculated.

[0099] Spatial candidate clusters that simultaneously meet the anomaly determination conditions are identified as anomalous node clusters.

[0100] Anomaly detection criteria include:

[0101] Condition 1: All detection nodes in the spatial candidate cluster are continuous anomaly detection nodes.

[0102] Condition 2: Calculate the spatial variance of the compensated gas concentration residuals for all detection nodes within each spatial candidate cluster. The difference between the expected variance when all detection nodes are considered as random discrete points and the spatial variance is greater than the difference threshold. The expected variance when all detection nodes are considered as random discrete points is the variance of the residuals collected and calculated from all detection nodes in the monitoring area under normal, leak-free conditions. The difference threshold is set based on experience and is mainly used to distinguish between the spatial variance when there is a leak and the expected variance when all detection nodes are considered as random discrete points. It can be adjusted according to the actual situation.

[0103] Condition 3: There is a reverse trajectory region, the value of L is greater than 0, and the Pearson correlation coefficient is greater than the correlation threshold; the correlation threshold is set based on empirical values, preferably 0.7-0.8.

[0104] A two-step method combining temporal continuous anomaly screening and spatial density clustering can accurately identify suspicious nodes that exhibit clustering in both time and space from a massive number of nodes, effectively eliminating random noise.

[0105] Methods for determining the location of a leak source include:

[0106] Obtain all abnormal node clusters, as well as the spatial coordinates and compensated gas concentration residual values ​​of all detection nodes within the abnormal node clusters.

[0107] Starting from the core node of the anomalous node cluster m, and driven by the current reverse wind direction and reverse wind speed (i.e., the wind direction is completely reversed), a reverse simulation is performed based on the Lagrange particle diffusion model. The source trajectory of the air mass in the past period is gradually traced back with a time step Δt2. The core node is the core point corresponding to the spatial candidate cluster of the anomalous node cluster when performing spatial clustering. Δt2 is less than Δt1. The time step Δt2 is set according to the actual situation, preferably 0.5 hours. The range of the upstream potential source region is obtained.

[0108] Extract all potential leakage sources whose spatial coordinates are located within the upstream potential source region to form an initial candidate source set for the anomalous node cluster m.

[0109] For each candidate source in the initial candidate source set, a forward simulation is performed based on the Lagrange particle diffusion model to calculate the theoretical concentration value at the spatial coordinates of each detection node in the anomalous node cluster m when the candidate source leaks. The Pearson correlation coefficient between the theoretical concentration value at the spatial coordinates of each detection node in the anomalous node cluster m and the actual observation residual value is calculated and normalized to obtain the spatial correlation score.

[0110] An exponential decay function is used to calculate the distance decay score based on the distance from the candidate source to the core node, such as... ,in, as candidate sources Distance decay score, as candidate sources Distance to the core node; This is a distance scale parameter, which can be set according to the monitoring range; distance When the distance is 0, the score is 1; as the distance increases, the score approaches 0. It is an exponential function with the natural constant as its base.

[0111] The spatial correlation score and distance attenuation score are weighted to obtain a comprehensive leakage score. The candidate source with the highest comprehensive leakage score is selected as the leakage source, and the spatial coordinates corresponding to the leakage source are taken as the leakage source location. The weighting weights can be set based on empirical values. The preferred weighting weight for spatial correlation is 0.7, and the preferred weighting weight for distance attenuation score is 0.3. These can be adjusted according to actual conditions.

[0112] By employing a two-stage physical verification process—reverse simulation for coarse localization of the source region followed by forward simulation for fine matching of the source points—and combining distance and correlation for comprehensive decision-making, the tracing results possess both physical rationality and quantitative credibility.

[0113] Example 2:

[0114] Please see Figure 4 As shown, this embodiment provides a laser detection system for gas leaks in a sealed environment, including:

[0115] Data acquisition module: It forms a sensor measurement network through detection nodes arranged in the monitoring area to collect the operational data, environmental data and monitoring data of the detection nodes;

[0116] The first analysis module takes the operational data, environmental data, and the pre-set digital geographic grid of the monitoring area as input to the pre-trained gas migration model to obtain the dynamic background concentration field distribution map of the entire monitoring area, which serves as the background concentration measurement benchmark at each spatial location.

[0117] The second analysis module matches and analyzes the spatial coordinates of each detection node with the dynamic background concentration field distribution map to obtain the local background prediction value of each detection node.

[0118] Concentration compensation module: Performs inversion analysis on monitoring data to obtain the original gas volume fraction measurement values ​​of each detection node; subtracts the corresponding local background prediction value from the original gas volume fraction measurement values ​​of each detection node to obtain the gas concentration residual value after compensation;

[0119] Leakage location module: Combines clustering algorithm to perform spatiotemporal analysis on the compensated gas concentration residual values ​​of all detection nodes in the entire monitoring area to obtain abnormal node clusters. Spatial matching and correlation analysis are performed on the abnormal node clusters with potential leak sources to obtain the location of the leak source.

[0120] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0121] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for laser detection of gas leaks for sealing environments, characterized in that, include: A sensor measurement network is formed by deploying detection nodes within the monitoring area to collect operational data, environmental data, and monitoring data from the detection nodes. The operational data, environmental data, and a pre-set digital geographic grid of the monitoring area are used as inputs to a pre-trained gas migration model to obtain a dynamic background concentration field distribution map of the entire monitoring area, which serves as a benchmark for background concentration measurement at each spatial location. The spatial coordinates of each detection node are matched and analyzed with the dynamic background concentration field distribution map to obtain the local background prediction value of each detection node. Inversion analysis of the monitoring data is performed to obtain the original gas volume fraction measurement values ​​of each detection node; the corresponding local background prediction value is subtracted from the original gas volume fraction measurement values ​​of each detection node to obtain the gas concentration residual value after compensation. By combining clustering algorithms to perform spatiotemporal analysis on the compensated gas concentration residual values ​​of all detection nodes in the entire monitoring area, abnormal node clusters are obtained. Spatial matching and correlation analysis are then performed on the abnormal node clusters and potential leakage sources to obtain the location of the leakage sources.

2. The laser detection method for gas leaks in a sealed environment according to claim 1, characterized in that, The operational data includes instantaneous injection / production flow rate, cavity pressure, and inventory change. The inventory change is obtained by collecting the cumulative inventory before and after injection / production and calculating the difference between the cumulative inventory before and after injection / production. The environmental data includes wind speed and wind direction. The monitoring data is the raw light intensity attenuation data of each detection node.

3. The laser detection method for gas leaks in a sealed environment according to claim 2, characterized in that, Methods for obtaining the original gas volume fraction measurements at each detection node include: The measurement optical path is corrected based on the reference optical path digital voltage sequence in the original light intensity attenuation data of each detection node, and common noise is eliminated to obtain the corrected light intensity sequence. The absorption rate as a function of time is calculated within each scanning cycle, and the optimal absorption spectral function is obtained by fitting the instantaneous wavelength of the laser. The optimal absorption spectral function is integrated and its ratio to the product of the monitoring optical path length and the absorption line intensity is calculated to obtain the measured value of the original gas volume fraction at each detection node.

4. The laser detection method for gas leaks in a sealed environment according to claim 3, characterized in that, Methods for obtaining the corrected light intensity sequence include: Within the current scanning cycle of the laser, select a spectral range; calculate the average values ​​of the digital voltage sequence of the measurement optical path and the digital voltage sequence of the reference optical path within the spectral range, respectively. The average values ​​of the digital voltage sequences of the measurement optical path and the reference optical path within the spectral range are subtracted from the digital voltage sequences of the measurement optical path and the reference optical path, respectively, to obtain the dominant sequence of the measured light intensity and the dominant sequence of the reference light intensity. At each sampling point, the ratio of the nominal value to the corresponding value of the dominant sequence of the reference light intensity is calculated as a correction factor. The product of each sequence value in the dominant sequence of the measured light intensity and the correction factor is calculated to obtain the corrected light intensity sequence.

5. The laser detection method for gas leaks in a sealed environment according to claim 1, characterized in that, Methods for obtaining anomalous node clusters include: By combining clustering algorithms, spatial clustering is performed on the compensated gas concentration residual values ​​of all detection nodes in the entire monitoring area to obtain G spatial candidate clusters; The detection node with the highest residual value of gas concentration after compensation in each spatial candidate cluster is taken as the central detection node, and the reverse trajectory simulation is performed in combination with environmental data; Check whether the reverse trajectory region passes through or points to L potential leakage sources; L is the number of potential leakage sources; if the value of L is greater than 0, then take the spatial coordinates of each corresponding potential leakage source as the center, perform a forward simulation according to the Lagrange particle diffusion model, calculate the theoretical concentration distribution corresponding to the spatial coordinates of each detection node in the spatial candidate cluster when the potential leakage source produces a leak; calculate the Pearson correlation coefficient between the theoretical concentration distribution corresponding to the spatial coordinates of each detection node in the spatial candidate cluster and the actual observation residual value. Spatial candidate clusters that simultaneously meet the anomaly determination conditions are identified as anomalous node clusters.

6. The laser detection method for gas leaks in a sealed environment according to claim 5, characterized in that, Methods for obtaining spatial candidate clusters include: For any detection node, calculate the mean and standard deviation of the compensated gas concentration residual value of the detection node within the sliding time window. If the mean is greater than three times the standard deviation, the corresponding detection node is judged to be abnormal within the current sliding time window. During the analysis period, if the number of consecutive sliding time windows in which a detection node is judged to be abnormal is greater than N or the cumulative proportion exceeds a preset ratio, the corresponding detection node is marked as a continuously abnormal detection node and recorded in the candidate detection node set. Based on the spatial coordinates of the detection node, a spatial clustering algorithm is used to cluster the detection nodes in the candidate detection node set to obtain G spatial candidate clusters, where k is a multiple of the standard deviation and N is the number of consecutive sliding time windows in which the detection node is judged to be abnormal.

7. The laser detection method for gas leaks in a sealed environment according to claim 1, characterized in that, Methods for determining the location of a leak source include: Obtain all abnormal node clusters, their corresponding detection node spatial coordinates, and the compensated gas concentration residual values; Starting from the core node of the abnormal node cluster m, a reverse simulation is performed based on the Lagrange particle diffusion model to trace the origin trajectory of the air mass; potential leakage sources located in the upstream potential source region are extracted to form an initial candidate source set. For each candidate source in the initial candidate source set, a forward simulation is performed to calculate the theoretical concentration value corresponding to each detection node in the anomaly node cluster m; the Pearson correlation coefficient between the theoretical concentration value corresponding to each detection node in the anomaly node cluster m and the actual observation residual value is calculated and normalized to obtain the spatial correlation score. An exponential decay function is used to calculate the distance decay score based on the distance from the candidate source to the core node; The spatial correlation score and distance decay score are weighted to obtain a comprehensive leakage score. The candidate source with the highest comprehensive leakage score is selected as the leakage source, and the corresponding spatial coordinates are output as the location of the leakage source.

8. The laser detection method for gas leaks in a sealed environment according to claim 1, characterized in that, In the preset digital geographic grid of the monitoring area, each grid point consists of spatial coordinates composed of the grid point's geographic coordinates and the corresponding elevation; the dynamic background concentration field distribution map provides the local background prediction value corresponding to each spatial coordinate point in grid form.

9. The laser detection method for gas leaks in a sealed environment according to claim 1, characterized in that, Methods for obtaining the local background prediction value for each detection node include: For each detection node, traverse each spatial coordinate point in the dynamic background concentration field distribution map. If the spatial coordinates of the detection node directly match the spatial coordinates of the point in the dynamic background concentration field distribution map, then use the local background prediction value corresponding to the spatial coordinates as the local background prediction value of the detection node; otherwise, obtain the spatial coordinates of the eight corner points of the grid cell to which the detection node belongs and the corresponding local background prediction values, and calculate the local background prediction value of the detection node through trilinear interpolation.

10. A laser detection system for gas leaks in a sealed environment, comprising the laser detection method for gas leaks in a sealed environment as described in any one of claims 1-9, characterized in that, include: Data acquisition module: It forms a sensor measurement network through detection nodes arranged in the monitoring area to collect the operational data, environmental data and monitoring data of the detection nodes; The first analysis module takes the operational data, environmental data, and the pre-set digital geographic grid of the monitoring area as input to the pre-trained gas migration model to obtain the dynamic background concentration field distribution map of the entire monitoring area, which serves as the background concentration measurement benchmark at each spatial location. The second analysis module matches and analyzes the spatial coordinates of each detection node with the dynamic background concentration field distribution map to obtain the local background prediction value of each detection node. Concentration compensation module: Performs inversion analysis on monitoring data to obtain the original gas volume fraction measurement values ​​of each detection node; subtracts the corresponding local background prediction value from the original gas volume fraction measurement values ​​of each detection node to obtain the gas concentration residual value after compensation; Leakage location module: Combines clustering algorithm to perform spatiotemporal analysis on the compensated gas concentration residual values ​​of all detection nodes in the entire monitoring area to obtain abnormal node clusters. Spatial matching and correlation analysis are performed on the abnormal node clusters with potential leak sources to obtain the location of the leak source.

Images