Sulfur hexafluoride concentration detection method based on dual-band TDLAS

Abstract

The invention belongs to the technical field of gas detection, and particularly relates to a sulfur hexafluoride (SF) concentration detection method and device based on a dual-band tunable diode laser absorption spectrum (TDLAS) and electronic equipment. In order to solve the problems that in an existing SFdetection technology, quantification and positioning functions are separated, the anti-interference capability is poor, and high sensitivity and rapid imaging cannot be synchronously achieved, the invention provides a method combining dual-band differential absorption and TDLAS. According to the scheme, a detection device in which an open type backward reflection path is combined with a built-in White cell (a multi-reflection cell) is adopted, 10.55 microns are selected as a measurement wave band, and 9 microns are selected as a reference wave band; through synchronous or alternate acquisition of dual-band signals, wavelet denoising preprocessing, differential absorption concentration inversion, Kalman filtering gas cloud tracking and Gaussian plume model leakage source inversion, rapid imaging positioning and accurate concentration quantification of an SFleakage point are realized. The system has the advantages of high detection sensitivity, strong anti-interference capability and integrated quantitative positioning function, and is suitable for efficient inspection and accurate diagnosis of the SFleakage of the power equipment.

Description

Technical Field

[0001] This invention belongs to the field of gas detection technology, specifically relating to a method for gas detection using tunable diode laser absorption spectroscopy (TDLAS) technology.

[0002] In particular, the present invention relates to a method and system for high-sensitivity, quantitative detection and imaging of sulfur hexafluoride (SF6) gas using dual-band infrared laser. Background Technology

[0003] Overview of existing SF6 detection technologies Sulfur hexafluoride (SF6) is a potent greenhouse gas with a global warming potential (GWP) approximately 23,500 times that of carbon dioxide, and an atmospheric lifetime of up to 3,200 years, exerting a significant long-term impact on climate change. Despite its extremely strong greenhouse effect, SF6 is widely used in the power industry due to its excellent insulation and arc-quenching properties, particularly as an insulating medium in high-voltage switchgear, gas-insulated switchgear, circuit breakers, and transformers. With the rapid expansion and aging of power infrastructure, SF6 leakage has become a significant problem. Leaks not only cause direct economic losses but also pose serious environmental hazards, exacerbating the global greenhouse effect. Therefore, developing efficient and accurate SF6 leak detection technologies is crucial for ensuring the safe operation of power systems, reducing environmental pollution, and complying with increasingly stringent environmental regulations. Effective leak detection helps maintenance personnel quickly locate leak points and promptly repair them, thereby minimizing SF6 emissions and environmental impact.

[0004] Currently, the main detection technologies for sulfur hexafluoride (SF6) gas include the following: (1). Electrochemical sensor method: The gas concentration is estimated by detecting the change in electrical signal generated by the chemical reaction between SF6 gas and the sensitive electrode.

[0005] (2). High-voltage breakdown method: Based on the insulation properties of SF6 gas, the purity of the gas or the leakage situation can be indirectly judged by measuring the change of breakdown voltage of the gas gap under high voltage.

[0006] (3). Vacuum decay method: The leakage rate is estimated by monitoring the rate of change of internal pressure over time after the sealed gas chamber is evacuated.

[0007] (4). Traditional single-band infrared imaging utilizes the absorption characteristics of SF6 gas in a specific infrared band (such as 10.6μm) to visualize and locate the leak point by capturing the absorption image of the gas cloud with an infrared camera.

[0008] (5) Single-band tunable diode laser absorption spectroscopy (TDLAS) technology measures the attenuation of the laser at the gas absorption peak and inversely calculates the gas concentration based on the Lambert-Beer law.

[0009] Deficiencies of existing technology: While existing technologies have been applied in certain scenarios, they all suffer from fundamental and insurmountable flaws, failing to meet the need for rapid, accurate, and comprehensive assessment of SF6 leaks. These flaws are mainly manifested in the following aspects: The functions of "quantitative" and "positioning" are separate and cannot be integrated: This is the most significant shortcoming of the existing technological system. While methods such as single-band TDLAS and electrochemical sensors can achieve quantitative measurement, they are all "point-based" detection. Their probes must be moved near the leak point to trigger an alarm, making wide-area scanning impossible and hindering rapid leak location. Inspection efficiency is extremely low, and they are prone to missing leaks in open areas or complex equipment structures. Traditional single-band infrared imaging can visualize and locate leak points, but its imaging principle relies on the temperature difference between the gas cloud and the background, making it a qualitative or semi-quantitative technique. It can only show "where the leak is," not "how much is leaking," and cannot provide precise concentration data and leakage rate, severely limiting its ability to assess leak severity and support subsequent handling decisions. Therefore, on-site maintenance must adopt the cumbersome process of "first using an infrared camera to locate (the leak), then using a contact probe to measure (quantitatively)," which is time-consuming and inconsistent.

[0010] Poor anti-interference ability, insufficient measurement accuracy and reliability: Electrochemical sensors are susceptible to interference from ambient temperature, humidity, and various cross-sensitive gases, leading to reading drift, low accuracy, and aging issues requiring frequent calibration. Traditional infrared imaging methods are highly susceptible to environmental factors, with imaging quality heavily reliant on the thermal radiation characteristics of the background. Imaging is impossible against a uniform background with little or no temperature difference, resulting in missed detections. Furthermore, solar reflection and other heat sources (such as steam) can easily cause false alarms. Single-band optical detection technologies (including single-band TDLAS and infrared imaging) are prone to interference from common paths such as optical path fluctuations, dust obstruction, and lens contamination, leading to distorted measurements and lacking effective online compensation mechanisms.

[0011] It is difficult to balance detection limit and sensitivity to meet application requirements: High-voltage breakdown and vacuum attenuation methods not only require the equipment to be powered off, but their detection sensitivity is also relatively low. They are usually used for coarse screening of large leaks and cannot detect early or small leaks. Single-band TDLAS, which can achieve high sensitivity, is limited by its point measurement mode and cannot quickly cover large areas.

[0012] In summary, the common deficiency of existing technologies lies in their inability to simultaneously achieve the two core functions of "visibility" (rapid imaging and positioning) and "accurate measurement" (high-precision quantitative analysis). The dual-band infrared TDLAS imaging detection method proposed in this invention aims to fundamentally solve this technical challenge. By utilizing the dual-band differential absorption principle, it achieves precise quantification by leveraging the high sensitivity and selectivity of TDLAS technology, while simultaneously enabling visualized positioning through image processing technology. Furthermore, the introduction of a reference band significantly suppresses common path interference, thereby achieving a leap in anti-interference capability and measurement accuracy in complex industrial environments. This invention aims to solve the problem of the separation between quantitative detection and rapid positioning functions in existing technologies by combining the dual-band differential absorption principle with TDLAS technology. Summary of the Invention

[0013] 1. Purpose of the invention: To overcome the aforementioned shortcomings of existing technologies, this paper provides a detection method capable of simultaneously and rapidly locating SF6 leak points and accurately quantifying leak concentrations. This method improves the detection's anti-interference capability, sensitivity, and accuracy. Ultimately, it constructs a complementary and capability-enhancing substation gas inspection solution.

[0014] 2. Technical Solution: Core concept: Employing the dual-band differential absorption principle combined with TDLAS technology, this method achieves imaging-based quantitative detection. By utilizing macroscopic scanning and microscopic confirmation, it resolves the contradictions between "efficiency" and "accuracy," and between "safety" and "effectiveness" in substation gas inspection.

[0015] Device Description: (1) Dual-band infrared TDLAS Dual-band TDLAS refers to a system that uses an infrared sensor, image processing circuitry, and motor controller to achieve dual-band imaging. It utilizes a filter wheel with a single detector or a beam splitter with dual detectors to achieve alternating or synchronous imaging. Regardless of the chosen approach, it should be able to control the rotation using a stepper motor, enabling both open-type and built-in pool detection systems to share a single dual-band TDLAS detection device. This reduces costs and facilitates maintenance.

[0016] (2) Open back reflection path and built-in White cell (multiple reflection cell) (open detection is performed by dual-band TDLAS).

[0017] An open back-reflection path refers to the pre-installation of multiple corner reflectors as fixed optical targets in key monitoring areas within a substation. When a robot inspects along a predetermined route, it stops at a pre-set point and automatically aligns its external TDLAS optical head with the corner reflector at that point, initiating TDLAS measurement and feeding back gas concentration information along the path. Using an open back-reflection path not only improves detection sensitivity (the optical path is twice the distance from the robot to the reflector) but also ensures the safety of the robot and equipment (detection can be performed while maintaining a safe distance from high-voltage equipment).

[0018] An internal White cell (multiple reflection cell) refers to a robot with a sampling port and a miniaturized White cell (multiple reflection cell) built in. It uses a micro pump to draw surrounding air into the internal gas absorption cell, measures the concentration of the inhaled gas after multiple reflections, and then discharges the internal gas to prepare for the next sampling.

[0019] The advantage of using a built-in White cell (multiple reflection cell) is that, since the measurement process is completely in a closed air chamber, it has high environmental robustness, and by designing the intracavity optical path, it can achieve extremely high detection sensitivity.

[0020] In summary, combining the two approaches achieves a better balance between the mobility of the inspection robot and the advantages of TDLAS telemetry. Open-loop detection handles large-area area scanning and early warning, while built-in pool detection handles precise measurement and leak point confirmation within a smaller area. Their complementary functions simultaneously improve inspection efficiency and detection accuracy. The system can acquire three-dimensional information including spatial distribution, concentration gradient, and leak source location. Furthermore, the two systems can mutually verify and back each other up, enhancing the overall reliability of the solution.

[0021] (3) Filter band selection: First band (measurement band): The strong absorption peak of SF6 gas at 10.55 μm was selected as the measurement channel.

[0022] Second band (reference band): A 9μm band, which is very close to the measurement band but where SF6 gas has almost no absorption, was selected as the reference channel.

[0023] Reasons for selection: Choosing 9μm as the reference band is based on considerations of spectral characteristics, hardware compatibility, and anti-interference capability. Spectrally, sulfur hexafluoride gas has an extremely low absorption cross-section in the 9μm band, which can be considered a weakly absorbing transparent window. It also has relatively few absorption lines for common atmospheric interfering gases such as H₂O and CO₂, allowing for the implementation of the main differential function while ensuring the stability of the reference signal itself. Hardware-wise, mid-infrared mercury cadmium telluride detectors typically have high responsivity in a wide spectral range of 8µm to 12µm, ensuring the reference channel receives a sufficiently strong and high-quality signal with a good signal-to-noise ratio. Furthermore, since both the measurement and reference bands are in the mid-infrared long-wave range, the same basic substrate material can be used, making fabrication easier. Regarding anti-interference capability, the appropriate distance (approximately 1.5μm) ensures that the absorption characteristics of sulfur hexafluoride are significantly different, while also ensuring that the effects of window contamination, dust scattering, minor errors from mechanical vibration, and background radiation fluctuations are highly consistent between the two.

[0024] Method and steps: Step 1: Construct an overall map of the substation, and pre-install multiple corner reflectors as fixed optical reflective targets in key monitoring areas within the substation.

[0025] Step 2: The inspection robot patrols along a fixed route. When it passes a preset corner reflector point, it automatically activates the open TDLAS to perform a "remote scan" of the route, controls the tunable diode laser (preferably QCL) to make its output wavelength alternately or simultaneously cover the measurement band and the reference band, and performs high-frequency wavelength modulation (WMS) on the laser.

[0026] Step 3: Dual-band signal acquisition: The laser beam passes through the area under test and is received by the photodetector. This is achieved through spatial multiplexing: a beam splitter divides the optical path into two beams, which pass through narrowband filters in the measurement and reference bands respectively, and are simultaneously acquired by two independent detectors.

[0027] Step 4: Preprocess the acquired dual-band raw signals to suppress Gaussian white noise and impulse noise introduced by environmental electromagnetic interference and the detector itself. For this purpose, we employ wavelet transform thresholding denoising. Specifically, considering the characteristics of the TDLAS second harmonic signal, we select the 'sym8' wavelet basis, perform a 5-level decomposition, and apply a heuristic thresholding (Sqtwolog) strategy for soft thresholding of the detail coefficients at each level. The advantage of this method is that it effectively preserves the steep edges in the signal, avoiding the edge blurring caused by traditional Fourier filtering during denoising, thus laying the foundation for subsequent accurate concentration inversion. Step 5: Concentration Inversion and Gas Region Segmentation: According to the Lambert-Beer Law, the gas absorbance A(λ) = -ln(It(λ) / I0(λ)) = K(λ) * C * L. In the dual-band system of this invention, we effectively eliminate common path interference by calculating the differential absorbance A_diff between the measurement band and the reference band, i.e., A_diff = A_measure - η * A_reference, where η is a correction coefficient determined based on the ratio of light intensity of the non-gas absorption portion of the two bands. This differential absorbance A_diff(x,y) is proportional to the concentration path product [C*L]_eff at each pixel position (x,y) in the image, i.e., [C*L]_eff(x,y) = K_system * A_diff(x,y), where K_system is a system constant obtained through standard gas calibration.

[0028] For open detection paths, the optical path length L is a known fixed value (twice the distance from the robot to the corner mirror), so the gas concentration C(x,y) = [C*L]_eff(x,y) / L can be directly calculated. For detection using a built-in White cell (multiple reflection cell), the optical path length L is the fixed design optical path length of the cell, and the concentration C is the uniform concentration of the gas being measured. This fully demonstrates that the combination of "dual-band differential" and "Lambert-Beer law" achieves interference-resistant quantitative analysis.

[0029] The output of this step is a two-dimensional concentration path integral distribution map and its corresponding binary gas cloud mask, which provides input for subsequent gas cloud tracing and source intensity inversion.

[0030] Step 6: Gas cloud trajectory tracking and state optimization based on Kalman filtering: After completing the binarization segmentation of the gas region, in order to achieve stable and continuous gas cloud tracking in the complex substation environment and suppress trajectory jitter and brief loss caused by sudden changes in light, dust obstruction or noise in single-frame images, this invention introduces the Kalman filtering algorithm to make optimal estimation of the motion state of the gas cloud.

[0031] 1. State-space modeling: To address the dynamic characteristics of SF6 leak clouds in video sequences, we define state and observation vectors suitable for this application: State vector (x) kThe model is defined as [c_x, c_y, v_x, v_y, A]ᵀ, where (c_x, c_y) are the pixel coordinates of the cloud's centroid, (v_x, v_y) are the cloud's velocity (pixels / frame) in the x and y directions, and A is the cloud's area (number of pixels). This model treats the cloud as a dynamic entity with a tendency for uniform motion and the ability to change its area, and is used to characterize the spatiotemporal dynamics of the cloud.

[0032] Observation vector (z) k ): Directly obtained from the difference processing results of each frame of the image, defined as [z_cx, z_cy, z_A]ᵀ, that is, the centroid coordinates and area measured in real time by image processing algorithms (such as contour finding, centroid calculation) for each frame.

[0033] 2. Innovative Applications and Model Adaptation: The key to this invention lies in the deep integration of Kalman filtering with the physical characteristics and application scenarios of gas detection, specifically reflected in: Adaptive process noise covariance (Q k Considering that cloud motion (such as that affected by wind field) may change non-uniformly, this invention does not employ a fixed process noise model. Instead, it designs an adaptive adjustment method for Q based on wind field estimates calculated from wind speed sensor readings or image optical flow. k The mechanism involves increasing the process noise when an increase in wind speed or a sudden change in wind direction is detected, enabling the filter to respond more quickly to changes in the cloud state.

[0034] Observation noise covariance (R k Dynamic setting of R: The observed noise is not fixed. We dynamically set R based on the signal-to-noise ratio (SNR) of the current image and the sharpness of the binarized cloud outline. k When the signal-to-noise ratio is low and the outline is blurry, the observation value of that frame is considered unreliable, and R is automatically increased. k This makes the Kalman filter more inclined to believe the model's predictions; conversely, when the observation quality is high, RA is reduced. k This makes them more confident in the current measurements.

[0035] The optimal state estimate is used to generate stable gas cloud motion trajectories and provides reliable input for subsequent leak source inversion. Step 7: Leakage Source Inversion: Based on the two-dimensional concentration path product [C*L]_eff(x,y) distribution obtained from open detection, combined with real-time wind speed u and wind direction obtained from meteorological sensors, a Gaussian plume model is used to invert the leakage source intensity and location. Specifically, the observed [C*L]_eff distribution is considered as the product of the concentration c predicted by the model at the corresponding location and an equivalent optical path. Using the Levenberg-Marquardt nonlinear least squares optimization algorithm, parameters such as source intensity Q, leakage source coordinates (Xs, Ys), and effective leakage height H in the model are iteratively adjusted to minimize the root mean square error between the model-predicted two-dimensional distribution and the observed [C*L]_eff distribution.

[0036] Furthermore, to adapt to the actual situation of numerous equipment and complex wind fields in substations, the standard Gaussian plume model was modified, and a near-field diffusion coefficient lookup table based on computational fluid dynamics (CFD) was introduced to more accurately describe the spatial variation of diffusion parameters σy and σz in the near field, thereby improving the inversion accuracy in complex environments.

[0037] The inversion includes two modes: in the rapid early warning mode, the sum of the differential gray values ​​of all pixels within the gas cloud mask is calculated as the leakage trend indicator; in the precise quantitative mode, the Levenberg-Marquardt optimization algorithm is used to fit the observed two-dimensional concentration path integral distribution with the prediction of the Gaussian plume model to inversely deduce the leakage rate Q and the location of the leakage source.

[0038] Step 8: The calculation module autonomously plans a path to the area with higher concentration. When it moves to the suspicious area, it rotates the filter and other components into the built-in pool detection device and manipulates the sampling rod or adjusts its own device to perform close-range, multi-point aspiration sampling.

[0039] Step 9: The calculation module integrates the "regional leakage cloud map" obtained from the open detection with the "precise leakage point coordinates and concentration" obtained from the built-in pool detection to generate a complete diagnostic report. Attached Figure Description

[0040] Figure 1 The following is a flowchart of the overall process of the method of this invention. A represents system initialization and map construction, B represents robot inspection and remote scanning, C represents dual-band signal excitation and acquisition, D represents signal preprocessing, E represents concentration inversion, F represents source intensity inversion, G represents path planning and precise sampling, H represents gas cloud tracking, and I represents data fusion and report generation.

[0041] Figure 2 Schematic diagram of the absorption spectrum of SF6 gas. Green indicates the measurement band and blue indicates the reference band.

[0042] Figure 3 Schematic diagram of wavelength modulation spectroscopy (WMS) technology.

[0043] Figure 4 The process of extracting the second harmonic (2f) signal. Detailed Implementation

[0044] The following detailed description, in conjunction with the accompanying drawings and specific embodiments, provides a method, apparatus, and electronic device for detecting sulfur hexafluoride concentration based on dual-band TDLAS according to the present invention. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0045] Example 1: Dual-band TDLAS detection using a filter wheel and a single detector: 1. System Hardware Configuration and Selection This embodiment uses a combination of a filter wheel and a single detector to achieve dual-band detection. The system hardware mainly includes: Laser: A continuous-wave output quantum cascade laser (QCL) is adopted. The model can be a quantum cascade laser with a center wavelength tunable to 10.55 μm. Its core performance parameters are: center wavelength tunable to 10.55 μm, output power greater than 100 mW, and high-frequency wavelength modulation function.

[0046] Filter wheel: The filter wheel is equipped with at least two narrowband filters. The first filter is for the measurement band, with a center wavelength of 10.55 μm and a full width at half maximum (FWHM) ≤ 0.1 μm; the second filter is for the reference band, with a center wavelength of 9.0 μm and an FWHM ≤ 0.1 μm. The filter wheel is driven by a stepper motor, with a positioning accuracy better than 0.1°.

[0047] Single detector: A mercury cadmium telluride detector cooled by liquid nitrogen, with a detection band covering 8-12 μm and a peak detectivity D> 1×10^10 cm·√Hz / W.

[0048] Control and processing unit: The core is an embedded processor, whose internal FPGA is used to implement high-speed motor control and data acquisition logic, and ARM core is used to run signal processing algorithms.

[0049] 2. Control timing and data acquisition process Controlling the timing is key to achieving alternating detection: 1. The processor sends instructions through the FPGA to drive the stepper motor to rotate the measurement band filter into the optical path.

[0050] 2. After the motor is in position, the FPGA synchronously triggers the laser modulation and data acquisition card to acquire the detector signal in this band at a sampling rate of at least 1 MSps for a duration of T (e.g., 10 ms).

[0051] 3. Once completed, the FPGA controls the motor to rotate the reference band filter into the optical path.

[0052] 4. Similarly, acquire the reference band signal for a duration of T.

[0053] 5. By repeating this process, each detection point will obtain a set of alternating measurement / reference band signal sequences.

[0054] 3. Signal Processing Algorithm Flow In a processor, the core steps and appendices of the software execution algorithm flow Figure 1 Consistent with the above.

[0055] The advantages of this embodiment are its relatively simple hardware structure, low cost, and avoidance of matching issues between dual detectors. The disadvantage is that the data acquisition frequency is limited due to the mechanical switching of the filters, making it suitable for scenarios where real-time detection requirements are not extremely stringent.

[0056] Example 2: Dual-band TDLAS detection using a beam splitter and dual detectors: 1. System Hardware Configuration and Selection This embodiment uses a combination of optical beam splitting and dual detectors to achieve simultaneous dual-band detection. The hardware is modified from Embodiment 1 as follows: Beam splitter: A broadband beam splitter using a single zinc selenide substrate, with a splitting ratio of approximately 50 / 50 in the 8-12 μm band.

[0057] Dual detectors: Two identical mercury cadmium telluride detectors are used, and their signal responses are kept consistent through subsequent calibration procedures. The two detectors are placed side by side on the temperature control module to maintain the consistency and stability of their response characteristics as much as possible.

[0058] 2. Synchronous Data Acquisition Process 1. The laser beam emitted from the laser is split into two beams of similar energy by a beam splitter.

[0059] 2. A beam of light passes through a fixedly installed measurement band filter and is received by detector 1.

[0060] 3. Another beam of light passes through a fixedly installed reference band filter and is received by detector two.

[0061] 4. The signals from the two detectors are acquired simultaneously by the synchronous data acquisition card, eliminating the need for mechanical switching and achieving true synchronous measurement.

[0062] 3. Dual-detector response correction method This is the core technical problem that this embodiment needs to solve. During power-on initialization or periodic self-calibration, the system executes the following calibration process: 1. Adjust the laser wavelength to a band where SF6 has no absorption (e.g., 9.5 μm).

[0063] 2. Collect the output signals of the two detectors at different light intensity levels.

[0064] 3. Using the response of detector one as a benchmark, a set of correction coefficients (including gain coefficient α and bias coefficient β) is calculated for each pixel or channel of detector two by least squares fitting, and a normalized correction model is established: `Detector1_signal = α*Detector2_signal + β.

[0065] 4. In the formal measurement, all the original signals of detector 2 are corrected in real time using the above correction coefficients before being used in subsequent calculations.

[0066] The advantages of this embodiment are that it has no moving parts, high reliability, and a fast data acquisition rate, making it suitable for dynamic leak monitoring with extremely high real-time requirements. The disadvantages are higher hardware costs and the need for additional calibration steps to ensure consistency between the two detector channels.

[0067] Example 3: Hardware and software collaborative work example This embodiment uses a complete inspection cycle as an example to illustrate the working process of the device of the present invention: 1. Initialization: The inspection robot is powered on and executes the dual-detector calibration process described in Example 2. Kalman filter initialization.

[0068] 2. Remote Scanning: The robot moves to the position of reflector No. 1. The main controller initiates open TDLAS measurement according to the preset program. The system synchronously acquires dual-band signals according to the process in Example 2, and generates an SF6 concentration path integral distribution map after processing.

[0069] 3. Gas cloud tracking and inversion: The processing unit detects the concentration anomaly, starts Kalman filtering to track the gas cloud trajectory, and inverts the leakage rate Q based on the "precise quantitative mode" in step 6, which is approximately 0.5 L / min, initially locating the leakage source near the XX circuit breaker.

[0070] 4. Precise Confirmation: The main controller then plans a path and drives the robot to move to the area below the XX circuit breaker. The robotic arm rotates the optical head into the built-in White cell (multiple reflection cell) optical path and starts the micro-pump to draw in air for sampling. The system switches to the built-in cell detection mode and measures a stable SF6 concentration of 5000 ppm at this point.

[0071] 5. Report Generation: The processing unit integrates the "regional leakage cloud map" generated by the open detection with the "precise concentration value" and "leak source coordinates" obtained by the built-in pool detection, and automatically generates a diagnostic report containing the leakage location, concentration, rate and severity level, which is then sent to the control center via wireless network.

[0072] The above three embodiments fully and completely demonstrate the various implementation methods and workflows of the present invention, proving the feasibility and superiority of its technical solutions.

Claims

1. A method for detecting concentration of sulfur hexafluoride based on dual-band TDLAS, characterized in that, The method comprises the following steps: 1.1 Constructing a map of the area to be detected, and arranging corner reflectors as fixed optical reflection targets in the key monitoring area; 1.2 Controlling the patrol robot to move along a preset path, and starting open TDLAS detection at the corner reflector points, and emitting infrared laser with alternating or synchronous coverage of the measurement waveband and the reference waveband; 1.3 Collecting double-waveband laser signals, pre-processing the collected signals, and adopting a wavelet transform threshold method to suppress noise; 1.4 Inverting a gas concentration path integral distribution based on the double-waveband differential absorption principle and the Lambert-Beer law; 1.5 Image segmentation and trajectory tracking are performed on the detected gas area, and an improved Kalman filtering algorithm is used to optimize the gas cloud motion state estimation; 1.6 Based on the concentration distribution and wind field information, the leakage source position and leakage rate are inverted by combining a modified Gaussian plume model, and a potential leakage source area is preliminarily inverted; 1.7 Controlling the robot to move to the suspicious area, and switching to the built-in White cell (multiple reflection cell) for close-range sampling and concentration confirmation; 1.8 Fusing the open detection and built-in cell detection data to generate a diagnosis report containing the leakage position, concentration, rate and severity level.

2. The method of claim 1, wherein, The measurement waveband is a strong absorption peak of SF6 at 10.55 μm, and the reference waveband is a 9 μm waveband in which the SF6 absorption is negligible.

3. The method of claim 1, wherein, The signal pre-processing adopts a wavelet transform threshold denoising method.

4. The method of claim 1, wherein, The improved Kalman filtering algorithm comprises: 4.1 The state vector is defined as [cx, cy, vx, vy, A]ᵀ, which respectively represents the gas cloud centroid coordinates, motion speed and area; 4.2 Process noise covariance Q k Adaptive adjustment of wind field estimates based on wind speed sensor or image optical flow method 4.3 Observation noise covariance R k Dynamic setting according to image signal-to-noise ratio and cloud of gas profile clarity.

5. The method of claim 1, wherein, The modified Gaussian plume model inversion comprises: 5.1 The Levenberg-Marquardt nonlinear least squares optimization algorithm is used to iteratively adjust the source strength Q, the leakage source coordinates and the effective height H; 5.2 A near-field diffusion coefficient lookup table based on computational fluid dynamics precalculation is introduced to correct the spatial variation of the diffusion parameters σy and σz.

6. A detection device for implementing the method according to any one of claims 1 to 5, characterized in that, It comprises: 6.1 A double-waveband TDLAS detection unit, comprising a quantum cascade laser, a filter wheel / splitter, a detector and a signal modulation circuit; 6.2 An open detection optical path, comprising corner reflectors arranged in the patrol area and a robot-mounted optical alignment mechanism; 6.3 A built-in White cell (multiple reflection cell) detection unit, comprising a miniature White cell (White cell (multiple reflection cell)), a gas pump and a sampling mechanism; 6.4 A control and processing unit, integrating a main controller, a motor driver, a signal processing module and a communication module; 6.5 A robot moving platform, carrying the detection unit and having autonomous navigation capability.

7. The device according to claim 6, wherein the double-waveband TDLAS detection unit adopts one of the following two implementation modes: 7.1 Mode one: filter wheel and single detector combination, waveband switching is realized by rotating the filter wheel through a stepping motor; 7.2 Mode two: splitter and double detector combination, double-waveband synchronous detection is realized through optical splitting.

8. The apparatus of claim 7, wherein, When the splitter and double detector combination is adopted, a double detector response correction module is further included, which is used to perform the following correction process: 8.1 Collect the output signal of two detectors at different light intensity in SF6 non-absorption waveband; 8.2 Take detector one as the reference, calculate the gain coefficient α and bias coefficient β of detector two by least square fitting method; 8.3 Establish the correction model: Detector1_signal = α*Detector2_signal + β; 8.4 Real-time correct the signal of detector two in formal measurement.

9. The apparatus of claim 6, wherein, The control and processing unit is configured to perform wavelet transform threshold denoising, differential absorption concentration inversion, improved Kalman filter trajectory tracking, modified Gaussian plume model leakage source inversion and data fusion operation.

10. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor implements the method of any one of claims 1-5 when executing the program.

Images