Gas leakage monitoring system and method based on spectral analysis

Abstract

The invention relates to the technical field of gas monitoring, and provides a spectrum analysis-based gas leakage monitoring system, which comprises a tunable laser source, an optical curtain forming device, a scanning optical sensor, a spectrum analysis module, a three-dimensional spectrum model construction module, a leakage traceability and diffusion prediction module, an intelligent early warning module and a snapshot type multispectral imaging unit, and an open optical path and a distributed multi-node layout are adopted. According to the invention, through the steps of optical screen scanning, spectral data acquisition and processing, three-dimensional dynamic model construction, intelligent analysis leakage determination, leakage traceability visualization and adaptive learning, precise monitoring, identification, traceability and early warning of gas leakage are realized; gas identification, leakage traceability and diffusion prediction are realized based on a three-dimensional dynamic model, deep learning and fluid mechanics, the monitoring effect is optimized by adopting distributed layout and self-adaptive learning, rapid disposal is assisted by graded early warning, and the accuracy, real-time performance and safety of gas leakage monitoring are greatly improved.

Description

Technical Field

[0001] This invention relates to the field of gas monitoring technology, specifically to a gas leak monitoring system and method based on spectral analysis. Background Technology

[0002] With the widespread use of natural gas in industrial production and residential life, safety accidents caused by gas leaks occur frequently, resulting not only in economic losses but also threatening lives and the environment. Current gas usage scenarios are becoming increasingly complex, such as urban natural gas pipeline valve wells and industrial gas storage areas, placing higher demands on the real-time performance, accuracy, and coverage of gas leak monitoring. Efficient monitoring technologies are urgently needed to ensure safe gas usage.

[0003] Existing gas leak monitoring technologies have many shortcomings. Some use point sensors, which have limited monitoring range and are prone to blind spots. Some can monitor gases, but it is difficult to accurately identify the gas type and quantify the concentration, and they have a weak ability to capture low-concentration leak signals. At the same time, most technologies cannot build a three-dimensional model of gas concentration, making it difficult to trace the leak point and predict the diffusion path. The early warning methods are also limited and cannot provide timely and efficient guidance for operation and maintenance. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a gas leak monitoring system and method based on spectral analysis, which solves the problems of limited monitoring range, low accuracy of gas identification and concentration quantification, weak capture of low-concentration leak signals, and inability to construct three-dimensional models to achieve leak source tracing and diffusion prediction in existing gas leak monitoring technologies.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a gas leak monitoring system based on spectral analysis, comprising:

[0006] A tunable laser source for emitting a scanning laser beam with a spectral range tunable, said spectral range covering at least one characteristic absorption peak of the target leaking gas;

[0007] An optical curtain forming apparatus includes one or more reflector arrays arranged around the perimeter of a monitoring area for reflecting the scanning laser beam to form one or more optical curtains in three-dimensional space.

[0008] A scanning optical sensor is used to receive laser signals that have passed through the optical veil and been reflected or scattered, and the sensor scans synchronously with the tunable laser source;

[0009] The spectral analysis module is used to analyze the spectral attenuation of the received laser signal and, based on semiconductor laser absorption spectroscopy, to identify and quantify the types and concentrations of gases passing through the optical curtain by comparing the absorption differences between the emitted and received spectra at specific wavelengths.

[0010] The three-dimensional spectral model construction module is used to fuse spectral data from different spatial locations with scanning angle and time information to construct a three-dimensional dynamic model of gas concentration distribution within the monitoring area.

[0011] The leak source tracing and diffusion prediction module is used to calculate the most likely location of the leak point based on the three-dimensional dynamic model, combined with the gas diffusion model and real-time environmental parameters, and to predict the diffusion path and rate of the gas cloud.

[0012] The intelligent early warning module is used to generate graded safety early warning information based on the type, concentration, leakage rate, and diffusion prediction results of the leaked gas.

[0013] Preferably, the tunable laser source is a broadband mid-infrared laser, whose emitted laser spectrum covers 3.2 μm to 3.4 μm, in order to perform high-sensitivity detection of the characteristic absorption peaks of methane gas.

[0014] Preferably, the reflector array in the optical curtain forming device is installed at a position and angle such that the formed optical curtain can partially surround or be adjacent to the key parts to be monitored, including but not limited to valves, flanges or pipe connections, forming an invisible gas permeation monitoring boundary.

[0015] Preferably, the system further includes a snapshot multispectral imaging unit, which includes an uncooled infrared focal plane array and a multispectral channel infrared objective lens integrated at its front end. Each optical channel of the objective lens corresponds to a narrowband filter, and the center wavelength corresponds to the strongest characteristic absorption band of one or more common leaked gases, which is used to simultaneously perform snapshot imaging on multiple discrete characteristic absorption bands to enhance the gas identification and imaging capabilities under complex mixed gases or background interference.

[0016] Preferably, the system adopts an open optical path and a distributed multi-node layout: it includes a central host chassis and multiple open long optical path optical units deployed at different locations in the monitoring area; the host chassis is equipped with an optical switch, which is connected to each optical unit through an optical fiber network to realize time-division cyclic detection of multiple monitoring optical paths in a region.

[0017] Preferably, a gas leak detection method based on spectral analysis includes the following steps:

[0018] S1. Optical curtain scanning steps: Control the tunable laser source and scanning optical sensor to synchronously scan the optical curtain area defined by the reflector array, and acquire the transmission or reflection spectral signals of a series of spatial sampling points;

[0019] S2. Spectral data acquisition and processing steps: During the scanning process, the raw spectral data of each sampling point is acquired synchronously, and the data is preprocessed using a scene-based non-uniformity correction algorithm to eliminate sensor noise and environmental background interference.

[0020] S3. Three-dimensional dynamic model construction steps: Associate the geographical coordinates, spectral absorption data and timestamps of each sampling point, and use spatial interpolation and data fusion algorithms to construct and update a three-dimensional dynamic spectral model that reflects the spatial distribution of gas concentration in real time;

[0021] S4. Intelligent Analysis and Leakage Determination Steps: Input the constructed three-dimensional spectral model data into the pre-trained deep learning model; this deep learning model is trained using a three-dimensional spectral database containing multiple gases, different concentrations, and interference scenarios, and is used for qualitative gas identification, quantitative concentration inversion, and authenticity determination of leak events;

[0022] S5. Leakage Source Tracing and Visualization Steps: After confirming the leak, based on the three-dimensional distribution of gas concentration and combined with the computational fluid dynamics inverse model, the location and intensity of the leak source are iteratively calculated; at the same time, the three-dimensional shape, concentration gradient and predicted diffusion path of the leaking gas cloud are fused and superimposed with the visible light video image to generate an augmented reality visualization alarm image.

[0023] Preferably, the deep learning model described in S4 is a convolutional neural network, whose input is a three-dimensional spectrum constructed with optical path length, wavenumber, and absorbance, and whose output is the type and concentration value of the target gas.

[0024] Preferably, the method further includes an adaptive learning step: the system will continuously collect new monitoring data and finally confirmed leakage event cases, and perform periodic incremental training and optimization on the deep learning model to continuously improve its ability to identify localized environmental characteristics and new leakage patterns.

[0025] Preferably, in S1, an optical path length adjustable detection method is adopted: by dynamically adjusting the scanning path or the angle of a specific unit in the reflector array, the effective absorption optical path of the laser in the suspected gas region is changed, thereby obtaining the absorption spectrum sequence of the same spatial point under different optical paths, which is used to enhance the detection signal-to-noise ratio of weak leakage signals at low concentrations.

[0026] Preferably, the specific algorithm for constructing the three-dimensional dynamic model in S3 includes:

[0027] This step executes a spatially correlated three-dimensional interpolation algorithm for spectral absorbance, which estimates the absorbance A of the target gas at the characteristic absorption peak wavelength λ at any unmeasured three-dimensional spatial point P(x,y,z) within the monitoring area. λ (P) is calculated through the following core steps:

[0028] Basic measurement: According to the Lambert-Beer law, at the i-th known measurement point S i At that location, the path integral absorbance A with respect to wavelength λ is obtained by laser spectral analysis. λ (S i )=-ln(I λ (S i ) / I 0λ );

[0029] Spatial interpolation modeling: Optimal unbiased estimation is performed using ordinary kriging, A λ (P) represents the linear weighted sum of the absorbance at n known measurement points in the surrounding area:

[0030]

[0031] Wherein, the weighting coefficient ω i It was determined by solving the Kriging equations, which are constructed from the spatial semivariograms between the measurement points;

[0032] Dynamic model update: The semi-variogram model is adaptively adjusted according to the spatial distribution characteristics of real-time measurement data, thereby realizing the dynamic and real-time update of the three-dimensional spectral model.

[0033] This invention provides a gas leak monitoring system and method based on spectral analysis. It has the following beneficial effects:

[0034] 1. This invention forms a multi-layer optical screen by using a tunable laser source and a reflector array, combined with a snapshot-type multispectral imaging unit, which can accurately capture low-concentration gas leak signals. Then, through the construction of a three-dimensional dynamic model and analysis of a deep learning model, it can achieve qualitative and quantitative identification of gas and determination of the authenticity of leaks. It can also be combined with a computational fluid dynamics reverse model to trace the source and predict diffusion, which greatly improves the monitoring accuracy and response efficiency.

[0035] 2. This invention adopts an open optical path and a distributed multi-node layout, combined with an adaptive learning step, which can perform time-division cyclic detection on multiple monitoring optical paths, continuously optimize the deep learning model, enhance the ability to identify localized environments and new leakage modes, and at the same time output information in multiple ways through hierarchical early warning, providing accurate basis for operation and maintenance personnel to handle the situation and reduce security risks. Attached Figure Description

[0036] Figure 1 This is a system block diagram of the present invention;

[0037] Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0038] 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.

[0039] Example:

[0040] Please see the appendix Figure 1 This invention provides a gas leak monitoring system based on spectral analysis, using methane leak monitoring in urban natural gas pipeline valve well areas as an application scenario. The system deployment includes the following:

[0041] Selection and installation of tunable laser sources:

[0042] A broadband mid-infrared laser was selected as the tunable laser source, with an emitted laser spectrum covering 3.2μm to 3.4μm, which fully encompasses the characteristic absorption peaks of methane gas. The laser was installed in a protective enclosure next to the valve well, which is waterproof and dustproof. The laser's emission angle was adjusted using a bracket to ensure that the laser beam was precisely directed towards the reflector array.

[0043] Layout of the optical curtain forming device:

[0044] The optical curtain forming device comprises four reflector arrays, positioned at the east, south, west, and north perimeters of the valve well. Each reflector array consists of ten high-reflectivity lenses, with the lens installation angles precisely adjusted: the east and west reflector arrays are tilted at a 45° angle, and the south and north reflector arrays are tilted at a 60° angle. This allows the scanning laser beam emitted by the tunable laser source to form three interlaced optical curtains within a 10-meter radius of the valve well after reflection. The optical curtains completely surround the valves, flanges, and pipe connections of the valve well, creating an invisible boundary for gas permeation monitoring.

[0045] Scanning optical sensor configuration:

[0046] A scanning optical sensor is installed inside a protective enclosure on the same side as the tunable laser source. The sensor's receiving lens is coaxial with the laser's output lens, and the sensor and laser are connected via a synchronization control line to achieve synchronous scanning. The sensor's signal receiving range covers the laser signal reflected or scattered by the optical curtain, and its sensitivity is set to capture the weak spectral attenuation signal caused by low concentrations of methane gas.

[0047] The spectral analysis module is integrated with the 3D spectral model construction module:

[0048] The spectral analysis module and the 3D spectral model construction module are integrated in a central host unit, which is placed in a monitoring room 50 meters away from the valve well. The spectral analysis module is connected to the scanning optical sensor via a data cable to receive the laser signal transmitted by the sensor. Based on semiconductor laser absorption spectroscopy, it compares the absorption difference between the emitted and received spectra at 3.3μm, which is the wavelength of the characteristic absorption peak of methane, to identify the gas type and quantify the concentration. The 3D spectral model construction module obtains the concentration data from the spectral analysis module, the scanning angle data from the scanning optical sensor, and the spatial location data from the GPS positioning module through a data interface to achieve multi-dimensional data fusion.

[0049] Leakage source tracing and diffusion prediction module parameter settings:

[0050] In the leak tracing and diffusion prediction module, a local meteorological parameter database is pre-imported, including wind speed, wind direction, temperature, humidity data for the past three years, and topographic data of the valve well area. The module's built-in gas diffusion model uses a Gaussian diffusion model, and is combined with computational fluid dynamics algorithms to ensure accurate inversion of the leak location and prediction of the diffusion path.

[0051] Intelligent early warning module configuration:

[0052] The intelligent early warning module works in conjunction with other modules within the central control unit, and presets three levels of safety warning thresholds: Level 1 warning is triggered when the methane concentration reaches 0-1% LEL (lower explosion limit); Level 2 warning is triggered when the concentration reaches 1%-5% LEL; and Level 3 warning is triggered when the concentration exceeds 5% LEL. Warning information is output through three methods: audible and visual alarms, SMS notifications, and a display screen in the monitoring room. The audible and visual alarms are installed around the valve well and in the monitoring room, while SMS notifications are sent to the mobile terminals of maintenance personnel.

[0053] Adding a snapshot-type multispectral imaging unit:

[0054] A snapshot-type multispectral imaging unit is installed on a utility pole 5 meters directly above the valve well. This unit includes an uncooled infrared focal plane array and a multispectral channel infrared objective lens at the front end. The infrared objective lens has four optical channels, each equipped with a narrowband filter. The center wavelengths of the filters correspond to the strongest characteristic absorption bands of methane (3.3μm), propane (3.4μm), ethane (3.5μm), and carbon monoxide (4.6μm), respectively. The imaging unit is connected to the central host unit via a network cable, and can simultaneously perform snapshot imaging of the four characteristic absorption bands, enhancing the ability to identify gases in complex gas mixtures or under background interference.

[0055] Implementation of open optical path and distributed multi-node layout:

[0056] The system adopts an open optical path and a distributed multi-node layout. In addition to the open long-path optical unit at the valve well, which includes a tunable laser source, a reflector array, and a scanning optical sensor, an identical open long-path optical unit is added every 100 meters along the natural gas pipeline to which the valve well belongs, for a total of 5 nodes. The central host box is equipped with an optical switch, which is connected to the 5 optical units through a single-mode fiber optic network. The optical switch switches the connection with each optical unit sequentially at a frequency of once every 30 seconds, so as to realize time-division cyclic detection of the 5 monitoring optical paths.

[0057] Please see the appendix Figure 2 This invention provides a gas leak monitoring method based on spectral analysis, comprising the following steps:

[0058] S1. Optical screen scanning procedure

[0059] Control commands are sent from the central control unit to activate the tunable laser source and scanning optical sensor. Both synchronously scan the optical curtain area defined by the reflector array at a speed of 30 revolutions per minute. During the scan, the laser beam sequentially passes through each spatial sampling point of the three layers of optical curtain. The scanning optical sensor simultaneously receives the laser signal after it passes through the optical curtain and is reflected, acquiring transmission spectral signals from a total of 200 spatial sampling points. The signal acquisition time for each sampling point is 0.1 seconds, ensuring complete capture of spectral information within the monitoring area.

[0060] Meanwhile, an adjustable optical path length detection method is adopted: when scanning to the sampling point directly above the valve in the valve well, the two reflector units on the east side of the reflector array are controlled by the central host box to adjust their angle from 45° to 50°, thereby changing the effective absorption optical path of the laser in the sampling point area from the original 5 meters to 8 meters. The absorption spectrum sequence of this spatial point under four different optical paths of 5 meters, 6 meters, 7 meters and 8 meters is obtained, thereby enhancing the signal-to-noise ratio of low-concentration weak leakage signals.

[0061] S2. Spectral Data Acquisition and Processing Steps

[0062] During the scanning process, the spectral analysis module of the central host chassis synchronously acquires raw spectral data from each sampling point. This raw data includes information such as wavelength, light intensity, and acquisition time. The acquired raw data is preprocessed using a scene-based non-uniformity correction algorithm: first, spectral data when the optical curtain does not detect gas is selected as a background reference value; then, the difference between the raw data at each sampling point and the background reference value is calculated to eliminate signal fluctuations caused by sensor noise. Simultaneously, temperature compensation is applied to the spectral data using real-time ambient temperature data obtained from a local temperature sensor to prevent environmental background interference from affecting subsequent analysis results. The preprocessed spectral data is stored in the central host chassis's database for later retrieval.

[0063] S3. Steps for Constructing a 3D Dynamic Model

[0064] The three-dimensional spectral model construction module retrieves preprocessed spectral data from the database and simultaneously obtains the geographical coordinates of each sampling point, provided by the GPS positioning module with an accuracy of 1 meter. These coordinates are synchronized with the timestamp and scanning time. The three are then correlated, and a three-dimensional interpolation algorithm based on spatial correlation of spectral absorbance is used to construct the model. The specific process is as follows:

[0065] Basic Measurements: According to the Lambert-Beer Law, at the i-th known measurement point Si, the path integral absorbance at a wavelength of 3.3 μm is obtained by laser spectral analysis as Aλ(Si) = -ln(Iλ(Di) / I0λ), where Iλ(Si) is the received light intensity at that measurement point, and I0λ is the emitted light intensity of the laser source. Calculations show that the absorbance at 200 measurement points ranges from 0.01 to 0.08.

[0066] Spatial interpolation modeling: For any unmeasured 3D spatial point P(x,y,z) within the monitoring area, the ordinary kriging method is used for optimal unbiased estimation. Its absorbance Aλ(P) is expressed as the linear weighted sum of the absorbances of the eight surrounding known measurement points, i.e., Aλ(P)=ω1·Aλ(S1)+ω2·Aλ(S2)+…+ω8·Aλ(S8). The weighting coefficients ωi are determined by solving the kriging equations, which are constructed from the spatial semivariograms among the eight known measurement points. A spherical model is selected for the semivariogram model, and the model parameters are determined by fitting the distance between the measurement points and the absorbance difference.

[0067] Dynamic model update: New spectral and spatial location data are collected every 5 minutes. Based on the spatial distribution characteristics of the new data, the parameters of the spatial semivariogram model are adaptively adjusted, and the absorbance of unmeasured points is recalculated to achieve real-time update of the three-dimensional spectral model. The updated model can clearly reflect the spatial distribution changes of methane concentration in the monitoring area.

[0068] S4. Intelligent Analysis and Leakage Detection Steps

[0069] The constructed three-dimensional spectral model data, including information on optical path length, wavenumber, and absorbance, is input into a pre-trained deep learning model. This deep learning model is a convolutional neural network, whose input layer receives a 32×32×32 three-dimensional spectral image, and whose output layer outputs the gas type and concentration value.

[0070] The training process of this convolutional neural network is as follows: A three-dimensional spectral database containing various gases such as methane, propane, ethane, and carbon monoxide, with a concentration range of 0-10% LEL, as well as interference scenarios such as rainy days, foggy days, and strong light, is pre-constructed. The database contains 100,000 sets of sample data. The network is trained using the stochastic gradient descent method with 500 training iterations and a learning rate of 0.001. The final model achieves a gas recognition accuracy of 98% and a concentration inversion error of less than 5%.

[0071] When the three-dimensional spectral model data is input, the convolutional neural network quickly analyzes it: if the output result is "Gas type: methane, concentration: 3.2% LEL", it is determined that there is a methane leak in the monitoring area; if the output result is "No target gas" or the concentration is less than 0.5% LEL, it is determined that there is no leak or that it is a background interference signal, thus realizing the determination of the authenticity of the leak event.

[0072] S5. Leakage Source Tracing and Visualization Steps

[0073] After confirming the presence of methane leakage, the leakage source tracing and diffusion prediction module used the concentration distribution data in the three-dimensional spectral model. The highest concentration was found at the sampling point directly above the valve, reaching 3.2% LEL, and the concentration decreased gradually towards the periphery. Combining the computational fluid dynamics inverse model, the location of the leakage source was iteratively calculated starting from the sampling point with the highest concentration. By adjusting the coordinate parameters of the leakage point in the model, the error between the concentration distribution simulated by the model and the actual monitored concentration distribution was minimized. Finally, the leakage point was determined to be located at the flange connection on the east side of the valve well, with a leakage intensity of 0.05 cubic meters per minute.

[0074] Simultaneously, the system integrates and overlays the following information: the three-dimensional morphology of the leaking gas cloud (three irregular spheres with concentrations of 0.5% LEL, 1% LEL, and 3% LEL as boundaries); the concentration gradient (a decrease of 0.3% LEL per meter from the leak point); the predicted diffusion path (based on a real-time wind speed of 2 m / s and a southwest wind direction, predicting that the gas cloud will diffuse 5 meters northeast after 10 minutes and 8 meters after 20 minutes); and visible light video footage of the valve well area (captured by a visible light camera mounted on a utility pole). This information generates an augmented reality visual alarm display on the monitor in the monitoring room. The leak point is marked in red, the concentration gradient is represented by different colored cloud maps, and the diffusion direction is indicated by arrows, allowing maintenance personnel to intuitively grasp the leak situation.

[0075] S6. Adaptive Learning Steps

[0076] During operation, the system continuously collects new monitoring data and confirmed leak incident cases. After each leak monitoring and handling is completed, the spectral data, concentration distribution data, leak point information, environmental parameters, and other data are compiled into a case and added to the training database of the deep learning model. The deep learning model undergoes periodic incremental training every 30 days, using a mixture of newly added case data and a portion of the original database to optimize the model's weight parameters and continuously improve its ability to identify localized environmental characteristics and novel leak patterns in the valve well area, such as minor flange leaks and aging valve seal leaks.

[0077] When the intelligent early warning module triggers a level-two warning based on the type, concentration, leakage rate, and diffusion prediction results of the leaked gas: the audible and visual alarm in the monitoring room emits an intermittent alarm sound, and the warning indicator light on the display screen turns yellow; simultaneously, the system automatically sends a warning SMS to the mobile phones of maintenance personnel, containing the leak location, leakage concentration, warning level, and handling suggestions. After receiving the warning information, maintenance personnel can understand the leak situation in advance through the visual alarm screen, quickly rush to the scene with professional equipment, effectively shorten the response time, and reduce safety risks.

[0078] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A gas leak monitoring system based on spectral analysis, characterized in that, include: A tunable laser source for emitting a scanning laser beam with a spectral range tunable, said spectral range covering at least one characteristic absorption peak of the target leaking gas; An optical curtain forming apparatus includes one or more reflector arrays arranged around the perimeter of a monitoring area for reflecting the scanning laser beam to form one or more optical curtains in three-dimensional space. A scanning optical sensor is used to receive laser signals that have passed through the optical veil and been reflected or scattered, and the sensor scans synchronously with the tunable laser source; The spectral analysis module is used to analyze the spectral attenuation of the received laser signal and, based on semiconductor laser absorption spectroscopy, to identify and quantify the types and concentrations of gases passing through the optical curtain by comparing the absorption differences between the emitted and received spectra at specific wavelengths. The three-dimensional spectral model construction module is used to fuse spectral data from different spatial locations with scanning angle and time information to construct a three-dimensional dynamic model of gas concentration distribution within the monitoring area. The leak source tracing and diffusion prediction module is used to calculate the most likely location of the leak point based on the three-dimensional dynamic model, combined with the gas diffusion model and real-time environmental parameters, and to predict the diffusion path and rate of the gas cloud. The intelligent early warning module is used to generate graded safety early warning information based on the type, concentration, leakage rate, and diffusion prediction results of the leaked gas.

2. The gas leak monitoring system based on spectral analysis according to claim 1, characterized in that, The tunable laser source is a broadband mid-infrared laser, whose emitted laser spectrum covers 3.2 μm to 3.4 μm, enabling high-sensitivity detection of the characteristic absorption peaks of methane gas.

3. The gas leak monitoring system based on spectral analysis according to claim 1, characterized in that, The reflector array in the optical curtain forming device is installed at a position and angle such that the formed optical curtain can partially surround or be adjacent to the key parts to be monitored. The key parts include, but are not limited to, valves, flanges or pipe connections, forming an invisible gas permeation monitoring boundary.

4. The gas leak monitoring system based on spectral analysis according to claim 1, characterized in that, The system also includes a snapshot multispectral imaging unit, which includes an uncooled infrared focal plane array and a multispectral channel infrared objective lens integrated at its front end. Each optical channel of the objective lens corresponds to a narrowband filter, and the center wavelength corresponds to the strongest characteristic absorption band of one or more common leaked gases. This is used to simultaneously perform snapshot imaging on multiple discrete characteristic absorption bands to enhance the gas identification and imaging capabilities under complex mixed gases or background interference.

5. A gas leak monitoring system based on spectral analysis according to claim 1, characterized in that: The system adopts an open optical path and a distributed multi-node layout: it includes a central main unit and multiple open long optical path optical units deployed in different locations in the monitoring area; the main unit is equipped with optical switches, which are connected to each optical unit through an optical fiber network to realize time-division cyclic detection of multiple monitoring optical paths in a region.

6. A gas leak detection method based on spectral analysis, using a gas leak detection system based on spectral analysis as described in any one of claims 1-5, characterized in that, Includes the following steps: S1. Optical curtain scanning steps: Control the tunable laser source and scanning optical sensor to synchronously scan the optical curtain area defined by the reflector array, and acquire the transmission or reflection spectral signals of a series of spatial sampling points; S2. Spectral data acquisition and processing steps: During the scanning process, the raw spectral data of each sampling point is acquired synchronously, and the data is preprocessed using a scene-based non-uniformity correction algorithm to eliminate sensor noise and environmental background interference. S3. Three-dimensional dynamic model construction steps: Associate the geographical coordinates, spectral absorption data and timestamps of each sampling point, and use spatial interpolation and data fusion algorithms to construct and update a three-dimensional dynamic spectral model that reflects the spatial distribution of gas concentration in real time; S4. Intelligent Analysis and Leakage Determination Steps: Input the constructed three-dimensional spectral model data into the pre-trained deep learning model; this deep learning model is trained using a three-dimensional spectral database containing multiple gases, different concentrations, and interference scenarios, and is used for qualitative gas identification, quantitative concentration inversion, and authenticity determination of leak events; S5. Leakage Source Tracing and Visualization Steps: After confirming the leak, based on the three-dimensional distribution of gas concentration and combined with the computational fluid dynamics inverse model, the location and intensity of the leak source are iteratively calculated; at the same time, the three-dimensional shape, concentration gradient and predicted diffusion path of the leaking gas cloud are fused and superimposed with the visible light video image to generate an augmented reality visualization alarm image.

7. The gas leak monitoring method based on spectral analysis according to claim 6, characterized in that, The deep learning model described in S4 is a convolutional neural network. Its input is a three-dimensional spectrum constructed with optical path length, wavenumber, and absorbance, and its output is the type and concentration value of the target gas.

8. The gas leak monitoring method based on spectral analysis according to claim 6, characterized in that, The method also includes an adaptive learning step: the system will continuously collect new monitoring data and finally confirmed leakage event cases, and perform periodic incremental training and optimization on the deep learning model to continuously improve its ability to identify localized environmental characteristics and new leakage patterns.

9. The gas leak monitoring method based on spectral analysis according to claim 6, characterized in that, In S1, an adjustable optical path length detection method is adopted: by dynamically adjusting the scanning path or the angle of a specific unit in the reflector array, the effective absorption optical path of the laser in the suspected gas region is changed, thereby obtaining the absorption spectrum sequence of the same spatial point under different optical paths, which is used to enhance the signal-to-noise ratio of the detection of weak leakage signals at low concentrations.

10. The gas leak monitoring method based on spectral analysis according to claim 1, characterized in that, The specific algorithm for constructing the three-dimensional dynamic model described in S3 includes: This step executes a spatially correlated three-dimensional interpolation algorithm for spectral absorbance, which estimates the absorbance A of the target gas at the characteristic absorption peak wavelength λ at any unmeasured three-dimensional spatial point P(x,y,z) within the monitoring area. λ (P) is calculated through the following core steps: Basic measurement: According to the Lambert-Beer law, at the i-th known measurement point S i At that location, the path integral absorbance A with respect to wavelength λ is obtained by laser spectral analysis. λ (S i )=-ln(I λ (S i ) / I 0λ ); Spatial interpolation modeling: Optimal unbiased estimation is performed using ordinary kriging, A λ (P) represents the linear weighted sum of the absorbance at n known measurement points in the surrounding area: Wherein, the weighting coefficient ω i It was determined by solving the Kriging equations, which are constructed from the spatial semivariograms between the measurement points; Dynamic model update: The semi-variogram model is adaptively adjusted according to the spatial distribution characteristics of real-time measurement data, thereby realizing the dynamic and real-time update of the three-dimensional spectral model.

Images