A method and system for detecting SF6 gas leaks based on photoacoustic spectroscopy.

Abstract

This invention discloses an SF6 gas leak detection method and system based on a photoacoustic spectroscopy analyzer, comprising: deploying sampling terminals to collect SF6 gas samples, obtaining stable gas input through constant current sampling and steady-state preprocessing; introducing the steady-state gas into a photoacoustic spectroscopy analyzer, where a self-tuning unit achieves resonant frequency stabilization and signal amplification output; performing dual-microphone differential detection and digital filtering to output a high signal-to-noise ratio stable SF6 detection signal; performing time calibration, anomaly removal, and consistency processing on the detection signal to generate standardized detection data; inputting the standardized data into a Transformer model to invert and generate the SF6 leak source location and diffusion path; displaying the inversion results on a monitoring interface, triggering an audible and visual alarm when exceeding the time limit, and uploading the data to a cloud monitoring platform. This invention achieves highly sensitive detection and accurate location of SF6 gas leaks by constructing an intelligent photoacoustic spectroscopy system combining photoacoustic-fluid steady-state control and self-tuning detection.

Description

Technical Field

[0001] This invention relates to the field of gas detection and intelligent monitoring technology, and in particular to a method and system for detecting SF6 gas leaks based on a photoacoustic spectroscopy analyzer. Background Technology

[0002] SF6 gas, as an excellent insulating and arc-quenching medium, is widely used in high-voltage electrical equipment such as circuit breakers, combined electrical appliances, gas-insulated busbars, and substations during the operation and maintenance of power equipment. However, SF6 is a strong greenhouse gas, and its leakage can not only lead to a decline in the insulation performance of equipment but also cause environmental pollution and safety hazards. Existing methods for detecting SF6 gas leaks mainly include infrared absorption, chemical sensing, and traditional photoacoustic spectroscopy. Although these methods can detect the gas, they still have several technical limitations in the complex environments of power equipment.

[0003] Currently, traditional photoacoustic spectroscopy detection systems generally employ a single-cavity fixed structure, which is significantly affected by temperature, humidity, and airflow disturbances. This leads to easy drift in the photoacoustic resonant frequency, resulting in insufficient detection sensitivity and signal stability. Simultaneously, pressure fluctuations and airflow noise are easily generated during gas sampling and fluid transmission, severely interfering with the photoacoustic signal quality and causing fluctuations in measurement results. Traditional algorithmic inversion models are mostly based on empirical formulas or simple linear regression, lacking constraints on the physical laws of gas diffusion, making it difficult to accurately locate and interpret the leak source. In existing detection systems, alarm thresholds are usually fixed and fail to adaptively adjust according to environmental changes or detection confidence levels, easily leading to false alarms or missed alarms. Current technologies cannot simultaneously achieve high stability, high accuracy, and intelligent response capabilities in the detection system.

[0004] Therefore, how to provide a method and system for detecting SF6 gas leaks based on photoacoustic spectroscopy is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a method and system for detecting SF6 gas leaks based on a photoacoustic spectroscopy analyzer. This invention comprehensively utilizes photoacoustic spectroscopy detection technology, intelligent sensing technology, and a physically interpretable Transformer deep learning algorithm to construct a complete detection system consisting of intelligent sampling, fluid steady-state control, self-tuned photoacoustic detection, signal differential processing, data synchronization, and leak inversion and alarm functions. By introducing a self-tuned detection mechanism using a double-sided composite photothermal response film in the detection stage, automatic frequency stabilization and signal amplification of the photoacoustic resonant frequency are achieved. In the sampling stage, an intelligent sensing sampling module and a photoacoustic-fluid isolation field control module are used to ensure the steady-state delivery of SF6 gas samples. In the data analysis stage, combined with a physically interpretable Transformer inversion model, high-precision identification and interpretable visualization analysis of the leak source location, leak intensity, and diffusion path are achieved. This invention possesses advantages such as high detection sensitivity, strong anti-interference capability, accurate leak location, and real-time intelligent response, enabling stable, accurate, and intelligent monitoring of SF6 gas leaks in complex power equipment environments.

[0006] An SF6 gas leak detection method based on a photoacoustic spectroscopy analyzer according to an embodiment of the present invention includes:

[0007] Intelligent sensor sampling terminals are deployed in the operating environment of power equipment to collect SF6 gas samples in the detected area. The collected SF6 gas samples are input into the photoacoustic-fluid isolation field control unit through a constant current sampling device for steady-state preprocessing to obtain steady-state SF6 gas samples.

[0008] A steady-state SF6 gas sample is introduced into the self-tuning detection unit of the photoacoustic spectrometer. The self-tuning detection unit includes a main detection cavity, a buffer cavity, and a double-sided composite photothermal response film layer set between the two cavities, and outputs a steady-state amplified SF6 photoacoustic signal.

[0009] The SF6 photoacoustic signal is subjected to dual-microphone differential detection through the signal acquisition module of the photoacoustic spectrum analyzer. After lock-in amplification, digital filtering and power normalization, a stable SF6 detection signal with high signal-to-noise ratio is obtained.

[0010] In the data acquisition and synchronization module, the stable SF6 detection signal is time-stamped, abnormal data is removed, and signal consistency is processed to generate a standardized detection dataset containing the spatial location of each sampling point, SF6 concentration value, and environmental parameters.

[0011] The standardized SF6 detection dataset is input into the leakage source inversion model based on a physically interpretable Transformer structure. The leakage source inversion model includes a physically constrained coding layer, a multi-scale attention fusion layer, and an interpretable inversion decoding layer to form the SF6 leakage source inversion result.

[0012] The SF6 leakage source inversion results are displayed on the visual monitoring interface. When the SF6 concentration value in the detection results exceeds the preset leakage concentration threshold, an alarm is triggered, an audible and visual alarm signal is output, and the SF6 leakage detection results and alarm information are uploaded to the cloud monitoring platform.

[0013] Optionally, the deployment of intelligent sensing sampling terminals in the power equipment operating environment refers to setting up several intelligent sensing sampling terminals in a closed or semi-closed space containing SF6 gas in a substation, circuit breaker bay, gas-insulated busbar, or cable corridor, according to the airflow direction, equipment layout, and leakage path. The intelligent sensing sampling terminals have built-in temperature, humidity, and airflow sensors and are connected to a constant current sampling device to synchronously collect SF6 gas samples and environmental parameters at different locations.

[0014] Optionally, the photoacoustic-fluid isolation field control unit is the front-end gas path module of the photoacoustic spectrometer, including an airflow isolation ring structure, a variable pressure field control plate, and a micro-circulation channel. The airflow isolation ring structure reduces airflow disturbance, the variable pressure field control plate maintains constant detection gas path pressure, and the micro-circulation channel maintains stable temperature, humidity, and density of SF6 gas, thereby obtaining a steady-state SF6 gas sample.

[0015] Optionally, the output steady-state amplified SF6 photoacoustic signal includes:

[0016] A steady-state SF6 gas sample is introduced into the main detection chamber and buffer chamber of the self-tuning detection unit. The buffer chamber is equipped with a partitioned pressure equalization chamber and a micro-grid impedance layer to perform pressure tuning and flow rate shaping on the SF6 gas entering the main detection chamber.

[0017] A mid-infrared modulated laser beam is introduced into the main detection cavity to form a fixed optical path. An annular variable optical path reflector is set on the inner wall of the main detection cavity to limit the round-trip path length of the beam. Temperature and pressure sensors are set on the outer wall of the main detection cavity to collect cavity state parameters.

[0018] A double-sided composite photothermal response film is set between the main detection cavity and the buffer cavity. One side of the double-sided composite photothermal response film faces the main detection cavity and is integrated with the annular micro-thermal driving electrode and temperature sensing unit. The other side faces the buffer cavity and is integrated with the micro-piezoelectric actuator. The surface reflection characteristics of the film are changed by the linkage between the temperature sensing unit and the micro-thermal driving electrode, and the spatial position of the film is finely adjusted by the micro-piezoelectric actuator.

[0019] During the linkage adjustment process, based on the changes in amplitude and phase of the acoustic signal collected in the main detection cavity and the changes in the state parameters of the main detection cavity and the buffer cavity, the ring variable optical path reflector, the micro-thermal driving electrode and the micro piezoelectric actuator are controlled to work together to output a stable SF6 photoacoustic raw signal.

[0020] The original SF6 photoacoustic signal is amplified and stabilized once in the self-tuning detection unit, and then packaged and output together with the state parameters of the main detection cavity and the buffer cavity to form a steady-state amplified SF6 photoacoustic signal.

[0021] Optionally, obtaining a stable SF6 detection signal with a high signal-to-noise ratio includes:

[0022] The SF6 photoacoustic raw signal is applied to a dual-microphone differential array set on both sides of the main detection cavity. The dual microphones are symmetrically installed with a half-wavelength spacing corresponding to the target acoustic resonance frequency, and are independently amplified by a phase-synchronized low-noise preamplifier driven by the same clock source.

[0023] Conductive protective rings and shielding layers are set around the dual microphones, and a miniature Helmholtz blind cavity is set between the dual microphones to attenuate flow noise. Primary cancellation of common-mode noise is achieved through hardware differential bridging with reverse-phase wiring.

[0024] The reference signal is obtained from the modulated laser drive end of the self-tuning detection unit. The phase-locked detection is performed on the electrical signal after differential bridging of the dual microphones to obtain the in-phase component and the quadrature component. Bandpass digital filtering and anti-aliasing processing are performed at a fixed sampling frequency. The filtering center is set to the frequency consistent with the acoustic resonance of the main detection cavity.

[0025] Adaptive weighting synthesis is performed on the filtered in-phase and quadrature components to form a differential amplitude signal. The differential amplitude signal is then normalized by the optical power monitoring channel built into the photoacoustic spectroscopy analyzer. The gain and reference phase of the preamplifier are automatically adjusted according to real-time environmental parameters.

[0026] Narrow-band calibration excitation is injected from the micro piezoelectric actuator in the buffer cavity at fixed time intervals. The dual-channel amplitude and phase registration coefficients are updated using the narrow-band calibration excitation without interfering with the measurement bandwidth. The normalized differential amplitude, phase, timestamp, and corresponding sampling end identifier are used as a stable SF6 detection signal with high signal-to-noise ratio.

[0027] Optionally, generating a standardized detection dataset containing the spatial location of each sampling point, SF6 concentration values, and environmental parameters includes:

[0028] It receives stable SF6 detection signals and corresponding timestamps, sampling end identifiers, status parameters of the main detection cavity and buffer cavity, temperature parameters, humidity parameters and airflow parameters, establishes a unified time axis based on the central control clock of the photoacoustic spectrometer, and compensates for the transmission delay and sampling jitter of each channel to generate the original record after time alignment.

[0029] The original records are subjected to anomaly data removal, including removing data segments corresponding to missing timestamps, abrupt amplitude changes, phase jumps, power monitoring mismatches, and sensor offline conditions. Short-term missing segments are filled by interpolation with adjacent valid segments to obtain valid records after anomaly removal.

[0030] Based on the identifiers of each sampling terminal and their installation coordinates in the power equipment environment, the valid records are mapped into spatial records, the sampling step size is unified, cross-channel amplitude and phase consistency correction and channel gain consistency calibration are completed, and the spatial records after consistency processing are output.

[0031] The spatial records are associated with temperature, humidity and airflow parameters from intelligent sensor sampling terminals. Standard entries are generated according to the sampling terminal number, and the units of measurement and range are standardized for each field to form a set of compliant entries.

[0032] The set of compliant items is cataloged and stored in chronological order and sampling order to generate a standardized test dataset containing the spatial location of each sampling point, SF6 concentration-related measurements, and environmental parameters.

[0033] Optionally, the process of generating the SF6 leakage source inversion result includes:

[0034] Receive standardized detection dataset, establish a spatiotemporal sequence according to sampling time order and sampling end spatial coordinates, extract SF6 concentration, airflow velocity direction, temperature, humidity and air pressure of each sampling point, and generate input item set;

[0035] In the physical constraint coding layer, the input entry set is coded in two levels. The first level of coding generates topological features based on the spatial topological relationship of the sampling points and the connectivity of the airflow channels. The second level of coding generates temporal features based on the joint changes of concentration, temperature and air pressure. The topological features and temporal features are constrained and fused based on the criterion of leakage accessibility to form a spatiotemporal feature sequence.

[0036] In the multi-scale attention fusion layer, a hierarchical attention structure of local diffusion sub-window and global propagation window is established. Features at different spatial locations within the same time period are fused horizontally, and features at the same spatial location at different times are fused vertically. The stability of environmental parameters is used as a weight to adaptively match the outputs of horizontal and vertical fusion to obtain feature representations.

[0037] In the interpretable inversion decoding layer, a candidate leak source grid is constructed and constrained by the device geometry boundary and airflow direction. The feature representation is then sparsified and inverted to output the three-dimensional coordinates of the leak source, the leak intensity estimate, and the diffusion path confidence. At the same time, interpretable visualization results containing diffusion direction, diffusion coverage, and boundary consistency are generated.

[0038] The interpretable visualization results are checked for consistency. The check includes consistency of multi-point concentration back-substitution, consistency of diffusion path and airflow direction, and consistency of equipment geometric boundary matching. If the check passes, the leakage source inversion result is output. If the check fails, the candidate leakage source mesh is automatically updated based on the feature representation and the inversion and check process is repeated until the result meets the consistency requirements.

[0039] Optionally, when the SF6 concentration value in the detection result exceeds a preset leakage concentration threshold, an alarm is triggered, and an audible and visual alarm signal is output, including:

[0040] The visualization monitoring interface displays the SF6 leakage source inversion results, records the display time, sampling end identifier and device location information, and generates event identifiers for local event registration.

[0041] The real-time detected SF6 concentration value is compared with the preset leakage concentration threshold and graded threshold set. When any threshold condition is met, the alarm module triggers an alarm and outputs an audible and visual alarm signal according to the corresponding threshold level. The alarm level and current confidence level are marked on the monitoring interface.

[0042] The event identifier, SF6 leakage source inversion results, confidence level, alarm level, concentration value, threshold parameter, time information and sampling end identifier are compiled into an alarm data frame and uploaded to the cloud monitoring platform through the communication link. After receiving the platform's confirmation, the data is archived locally.

[0043] An SF6 gas leak detection system based on a photoacoustic spectroscopy analyzer according to an embodiment of the present invention includes the following modules:

[0044] The intelligent sampling module is used to deploy multiple sampling points in the power equipment environment to collect SF6 gas samples in real time.

[0045] The fluid steady-state control module is used to perform steady-state preprocessing on SF6 gas samples and output steady-state SF6 gas samples.

[0046] The self-tuning detection module is used to import steady-state SF6 gas samples into the photoacoustic spectrometer for detection and output steady-state amplified SF6 photoacoustic signals.

[0047] The signal differential processing module is used to perform dual-microphone differential detection on the SF6 photoacoustic signal, and obtain a stable SF6 detection signal through lock-in amplification, digital filtering and power normalization.

[0048] The data synchronization module is used to perform time calibration, anomaly removal, and consistency processing on stable SF6 detection signals to generate a standardized detection dataset.

[0049] The leakage inversion and alarm module is used to input standardized data into the leakage source inversion model for inversion analysis and display the results on the monitoring interface. When the concentration exceeds the limit, it triggers an audible and visual alarm and uploads the data to the cloud monitoring platform.

[0050] The beneficial effects of this invention are:

[0051] This invention effectively solves the problem of decreased detection sensitivity in existing photoacoustic spectroscopy detection systems due to airflow disturbances, temperature and humidity changes, and pressure fluctuations in complex power equipment environments by introducing photoacoustic-fluid isolation field control technology and a self-tuning detection mechanism. Through the coordinated operation of the intelligent sensing sampling module and the constant current sampling device, steady-state control of SF6 gas samples is achieved throughout the sampling, transportation, and detection processes, reducing interference from external environmental factors on the photoacoustic signal and enabling the system to maintain high stability and repeatability during long-term continuous operation.

[0052] The self-tuning detection module used in this invention innovatively sets up a main detection cavity, a buffer cavity, and a double-sided composite photothermal response film. Through the linkage of acoustic and photothermal feedback, it achieves self-stabilization of the resonant frequency and signal amplification, breaking through the limitations of traditional single-cavity structures that are sensitive to temperature drift and mechanical vibration. It can automatically maintain the optimal resonant state without manual calibration, improve the amplitude response and signal-to-noise ratio of the photoacoustic signal, and enable SF6 concentration changes to be captured more accurately.

[0053] This invention introduces a physically interpretable Transformer inversion model in the data analysis and result processing stages, achieving high-precision localization of SF6 leak sources and reconstruction of diffusion paths. The inversion model integrates physical constraints and a multi-scale attention mechanism, maintaining the expressive power of deep models while ensuring the physical rationality and interpretability of the results. Combined with visual monitoring and audible / visual alarm mechanisms, this invention can provide immediate alarms and cloud-based information linkage when the leak concentration exceeds a threshold, ultimately achieving high sensitivity, high reliability, and intelligent closed-loop control of the detection process, improving the safety and efficiency of SF6 gas leak detection. Attached Figure Description

[0054] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0055] Figure 1 This is a flowchart of an SF6 gas leak detection method based on a photoacoustic spectroscopy analyzer proposed in this invention;

[0056] Figure 2 This is a schematic diagram of the SF6 gas leak detection system based on a photoacoustic spectroscopy analyzer proposed in this invention. Detailed Implementation

[0057] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0058] refer to Figure 1 A method for detecting SF6 gas leaks based on photoacoustic spectroscopy analyzer, comprising:

[0059] Intelligent sensor sampling terminals are deployed in the operating environment of power equipment to collect SF6 gas samples in the detected area. The collected SF6 gas samples are input into the photoacoustic-fluid isolation field control unit through a constant current sampling device for steady-state preprocessing to obtain steady-state SF6 gas samples.

[0060] A steady-state SF6 gas sample is introduced into the self-tuning detection unit of the photoacoustic spectrometer. The self-tuning detection unit includes a main detection cavity, a buffer cavity, and a double-sided composite photothermal response film layer set between the two cavities, and outputs a steady-state amplified SF6 photoacoustic signal.

[0061] The SF6 photoacoustic signal is subjected to dual-microphone differential detection through the signal acquisition module of the photoacoustic spectrum analyzer. After lock-in amplification, digital filtering and power normalization, a stable SF6 detection signal with high signal-to-noise ratio is obtained.

[0062] In the data acquisition and synchronization module, the stable SF6 detection signal is time-stamped, abnormal data is removed, and signal consistency is processed to generate a standardized detection dataset containing the spatial location of each sampling point, SF6 concentration value, and environmental parameters.

[0063] The standardized SF6 detection dataset is input into the leakage source inversion model based on a physically interpretable Transformer structure. The leakage source inversion model includes a physically constrained coding layer, a multi-scale attention fusion layer, and an interpretable inversion decoding layer to form the SF6 leakage source inversion result.

[0064] The SF6 leakage source inversion results are displayed on the visual monitoring interface. When the SF6 concentration value in the detection results exceeds the preset leakage concentration threshold, an alarm is triggered, an audible and visual alarm signal is output, and the SF6 leakage detection results and alarm information are uploaded to the cloud monitoring platform.

[0065] In this embodiment, the deployment of intelligent sensing sampling terminals in the operating environment of power equipment refers to setting up several intelligent sensing sampling terminals in the closed or semi-closed space where SF6 gas exists in substations, circuit breaker bays, gas-insulated busbars, or cable corridors, according to the airflow direction, equipment layout, and leakage path. The intelligent sensing sampling terminals have built-in temperature, humidity, and airflow sensors and are connected to a constant current sampling device to synchronously collect SF6 gas samples and environmental parameters at different locations.

[0066] In this embodiment, the photoacoustic-fluid isolation field control unit is the front-end gas path module of the photoacoustic spectrometer, including an airflow isolation ring structure, a variable pressure field control board, and a micro-circulation channel. The airflow isolation ring structure reduces airflow disturbance, the variable pressure field control board maintains a constant detection gas path pressure, and the micro-circulation channel keeps the temperature, humidity, and density of SF6 gas stable, thereby obtaining a steady-state SF6 gas sample.

[0067] In this embodiment, the output of the steadily amplified SF6 photoacoustic signal includes:

[0068] A steady-state SF6 gas sample is introduced into the main detection chamber and buffer chamber of the self-tuning detection unit. The buffer chamber is equipped with a partitioned pressure equalization chamber and a micro-grid impedance layer to perform pressure tuning and flow rate shaping on the SF6 gas entering the main detection chamber.

[0069] A mid-infrared modulated laser beam is introduced into the main detection cavity to form a fixed optical path. An annular variable optical path reflector is set on the inner wall of the main detection cavity to limit the round-trip path length of the beam. Temperature and pressure sensors are set on the outer wall of the main detection cavity to collect cavity state parameters.

[0070] A double-sided composite photothermal response film is set between the main detection cavity and the buffer cavity. One side of the double-sided composite photothermal response film faces the main detection cavity and is integrated with the annular micro-thermal driving electrode and temperature sensing unit. The other side faces the buffer cavity and is integrated with the micro-piezoelectric actuator. The surface reflection characteristics of the film are changed by the linkage between the temperature sensing unit and the micro-thermal driving electrode, and the spatial position of the film is finely adjusted by the micro-piezoelectric actuator.

[0071] During the linkage adjustment process, based on the changes in amplitude and phase of the acoustic signal collected in the main detection cavity and the changes in the state parameters of the main detection cavity and the buffer cavity, the ring variable optical path reflector, the micro-thermal driving electrode and the micro piezoelectric actuator are controlled to work together to output a stable SF6 photoacoustic raw signal.

[0072] The original SF6 photoacoustic signal is amplified and stabilized once in the self-tuning detection unit, and then packaged and output together with the state parameters of the main detection cavity and the buffer cavity to form a steady-state amplified SF6 photoacoustic signal.

[0073] In this embodiment, obtaining a stable SF6 detection signal with a high signal-to-noise ratio includes:

[0074] The SF6 photoacoustic raw signal is applied to a dual-microphone differential array set on both sides of the main detection cavity. The dual microphones are symmetrically installed with a half-wavelength spacing corresponding to the target acoustic resonance frequency, and are independently amplified by a phase-synchronized low-noise preamplifier driven by the same clock source.

[0075] Conductive protective rings and shielding layers are set around the dual microphones, and a miniature Helmholtz blind cavity is set between the dual microphones to attenuate flow noise. Primary cancellation of common-mode noise is achieved through hardware differential bridging with reverse-phase wiring.

[0076] The reference signal is obtained from the modulated laser drive end of the self-tuning detection unit. The phase-locked detection is performed on the electrical signal after differential bridging of the dual microphones to obtain the in-phase component and the quadrature component. Bandpass digital filtering and anti-aliasing processing are performed at a fixed sampling frequency. The filtering center is set to the frequency consistent with the acoustic resonance of the main detection cavity.

[0077] Adaptive weighting synthesis is performed on the filtered in-phase and quadrature components to form a differential amplitude signal. The differential amplitude signal is then normalized using the optical power monitoring channel built into the photoacoustic spectroscopy analyzer. The preamplifier gain and reference phase are automatically adjusted based on real-time environmental parameters. Specifically, the adaptive weighting synthesis of the filtered in-phase and quadrature components involves:

[0078] The filtered in-phase and quadrature components are extracted separately, and their respective weighting coefficients are dynamically calculated based on the amplitude and phase characteristics.

[0079] Based on the current environmental noise level, signal stability, and amplitude-phase deviation, the weight allocation ratio of in-phase and quadrature components is adjusted in real time.

[0080] The in-phase and quadrature components are weighted and summed according to the allocation ratio, and the differential amplitude signal after adaptive fusion is output.

[0081] A narrow-band calibration excitation is injected at fixed time intervals from a micro-piezoelectric actuator in the buffer cavity. The narrow-band calibration excitation updates the amplitude and phase registration coefficients of the dual channels without interfering with the measurement bandwidth. The normalized differential amplitude, phase, timestamp, and corresponding sampling end identifier are used as a stable SF6 detection signal with high signal-to-noise ratio. The narrow-band calibration excitation refers to the mechanical excitation signal with a narrow spectral range and a clear center frequency generated periodically by the micro-piezoelectric actuator deployed in the buffer cavity according to the set frequency and amplitude. The excitation frequency is located at the edge or outside the measurement bandwidth. The mechanical excitation signal only generates energy distribution within local frequency points. It is used to perform amplitude and phase registration of the detection channel during calibration without interfering with or covering the actual SF6 gas measurement signal.

[0082] In this embodiment, generating a standardized detection dataset containing the spatial location of each sampling point, SF6 concentration values, and environmental parameters includes:

[0083] It receives stable SF6 detection signals and corresponding timestamps, sampling end identifiers, status parameters of the main detection cavity and buffer cavity, temperature parameters, humidity parameters and airflow parameters, establishes a unified time axis based on the central control clock of the photoacoustic spectrometer, and compensates for the transmission delay and sampling jitter of each channel to generate the original record after time alignment.

[0084] The original records are subjected to anomaly data removal, including removing data segments corresponding to missing timestamps, abrupt amplitude changes, phase jumps, power monitoring mismatches, and sensor offline conditions. Short-term missing segments are filled by interpolation with adjacent valid segments to obtain valid records after anomaly removal.

[0085] Based on the identifiers of each sampling terminal and their installation coordinates in the power equipment environment, the valid records are mapped to spatial records, the sampling step size is unified, cross-channel amplitude and phase consistency correction and channel gain consistency calibration are completed, and the spatial records after consistency processing are output, wherein:

[0086] Cross-channel amplitude and phase consistency correction refers to using a unified reference standard to normalize the amplitude and phase of photoacoustic detection signals acquired from different sampling channels. Through amplitude normalization and phase compensation, the amplitude and phase deviations caused by differences in hardware characteristics, wiring length, or time delay between channels are eliminated, so that the amplitude and phase characteristics of the signals of all channels are comparable under the same reference.

[0087] Channel gain consistency calibration refers to calibrating the response curves of the signal amplifiers of all sampling channels. Using a standard signal input, the output gain of each channel is measured, and the channel gain parameters are adjusted or corrected based on the actual measured gain differences to ensure that the output amplitude of each channel is consistent under the same input conditions, thus ensuring the uniformity and accuracy of the multi-channel signal strength.

[0088] The spatial records are associated with temperature, humidity and airflow parameters from intelligent sensor sampling terminals. Standard entries are generated according to the sampling terminal number, and the units of measurement and range are standardized for each field to form a set of compliant entries.

[0089] The set of compliant items is cataloged and stored in chronological order and sampling order to generate a standardized test dataset containing the spatial location of each sampling point, SF6 concentration-related measurements, and environmental parameters.

[0090] In this embodiment, the process of generating the SF6 leakage source inversion result includes:

[0091] Receive standardized detection dataset, establish a spatiotemporal sequence according to sampling time order and sampling end spatial coordinates, extract SF6 concentration, airflow velocity direction, temperature, humidity and air pressure of each sampling point, and generate input item set;

[0092] In the physical constraint coding layer, the input entry set is coded in two levels. The first level of coding generates topological features based on the spatial topological relationship of the sampling points and the connectivity of the airflow channels. The second level of coding generates temporal features based on the joint changes of concentration, temperature and air pressure. The topological features and temporal features are constrained and fused based on the criterion of leakage accessibility to form a spatiotemporal feature sequence.

[0093] In the multi-scale attention fusion layer, a hierarchical attention structure of local diffusion sub-windows and global propagation windows is established. Features from different spatial locations within the same time period are fused laterally, and features from the same spatial location at different times are fused vertically. The stability of environmental parameters is used as a weight to adaptively balance the outputs of the lateral and vertical fusions, resulting in a feature representation, where:

[0094] Lateral fusion of features from different spatial locations within the same time period refers to integrating SF6 concentration values ​​and corresponding environmental features obtained from different sampling points at the same time. By comparing and weighting the spatial distribution features, spatial correlation and local anomaly change information of leakage diffusion are extracted to achieve synergistic expression of spatial distribution features.

[0095] Vertical fusion of features at the same spatial location at different times refers to serializing the SF6 concentration values ​​and environmental features collected at the same sampling point at multiple consecutive time steps. By comparing and weighting the time-series features, the dynamic trends of gas concentration and environmental parameters over time can be captured, thereby achieving in-depth modeling of time-series evolution features and improving the temporal sensitivity of abnormal changes.

[0096] In the interpretable inversion decoding layer, a candidate leak source grid is constructed and constrained by the device geometry boundary and airflow direction. The feature representation is then subjected to sparse inversion, outputting the three-dimensional coordinates of the leak source, the leak intensity estimate, and the diffusion path confidence. Simultaneously, interpretable visualization results containing diffusion direction, diffusion coverage, and boundary consistency are generated. Specifically, the sparse inversion of the feature representation involves:

[0097] Based on the spatial distribution of equipment and the coordinates of sampling points, the monitoring area is divided into a finite number of three-dimensional candidate grids, and a sparse distribution assumption is established for each grid cell.

[0098] Based on multi-point concentration characteristics and environmental parameters, the feature representation is mapped to candidate grid cells, and the leakage intensity distribution of the main active grid cells is obtained through optimization strategies;

[0099] In the candidate grid sparse inversion, the sparse solution results that are consistent with the actual physical environment are selected by combining the device geometric boundary and airflow guidance constraints.

[0100] The interpretable visualization results are checked for consistency. The check includes consistency of multi-point concentration back-substitution, consistency of diffusion path and airflow direction, and consistency of equipment geometric boundary matching. If the check passes, the leakage source inversion result is output. If the check fails, the candidate leakage source mesh is automatically updated based on the feature representation and the inversion and check process is repeated until the result meets the consistency requirements.

[0101] In this embodiment, the step of triggering an alarm and outputting an audible and visual alarm signal when the SF6 concentration value in the detection result exceeds a preset leakage concentration threshold includes:

[0102] The visualization monitoring interface displays the SF6 leakage source inversion results, records the display time, sampling end identifier and device location information, and generates event identifiers for local event registration.

[0103] The real-time detected SF6 concentration value is compared with the preset leakage concentration threshold and graded threshold set. When any threshold condition is met, the alarm module triggers an alarm and outputs an audible and visual alarm signal according to the corresponding threshold level. The alarm level and current confidence level are marked on the monitoring interface.

[0104] The event identifier, SF6 leakage source inversion results, confidence level, alarm level, concentration value, threshold parameter, time information and sampling end identifier are compiled into an alarm data frame and uploaded to the cloud monitoring platform through the communication link. After receiving the platform's confirmation, the data is archived locally.

[0105] An SF6 gas leak detection system based on a photoacoustic spectroscopy analyzer according to an embodiment of the present invention includes the following modules:

[0106] The intelligent sampling module is used to deploy multiple sampling points in the power equipment environment to collect SF6 gas samples in real time.

[0107] The fluid steady-state control module is used to perform steady-state preprocessing on SF6 gas samples and output steady-state SF6 gas samples.

[0108] The self-tuning detection module is used to import steady-state SF6 gas samples into the photoacoustic spectrometer for detection and output steady-state amplified SF6 photoacoustic signals.

[0109] The signal differential processing module is used to perform dual-microphone differential detection on the SF6 photoacoustic signal, and obtain a stable SF6 detection signal through lock-in amplification, digital filtering and power normalization.

[0110] The data synchronization module is used to perform time calibration, anomaly removal, and consistency processing on stable SF6 detection signals to generate a standardized detection dataset.

[0111] The leakage inversion and alarm module is used to input standardized data into the leakage source inversion model for inversion analysis and display the results on the monitoring interface. When the concentration exceeds the limit, it triggers an audible and visual alarm and uploads the data to the cloud monitoring platform.

[0112] Example 1:

[0113] To verify the feasibility of this invention in practice, it was applied to a 110kV intelligent switchyard. The on-site detection environment was complex, with compact equipment, uneven gas flow, and frequent temperature and humidity changes. Traditional infrared absorption detectors and chemical sensors are prone to signal drift and false alarms in this environment, with an average false alarm rate of 9.8%, and they exhibit slow response to low-concentration SF6 leaks. This invention, through multi-point deployment of intelligent sensing sampling terminals, photoacoustic-fluid isolation steady-state control, a self-tuning detection module, and a leak inversion model based on a physically interpretable Transformer, achieves stable, high-precision detection and intelligent alarm for SF6 leaks in complex power environments.

[0114] During the on-site installation phase, the system deployed six intelligent sensor sampling terminals in the gas-insulated switchgear bays, the top of the circuit breaker gas chambers, busbar channels, and joint areas. Each sampling terminal was approximately 2 meters apart, with a sampling frequency of 1Hz, and data transmission utilized an industrial Ethernet link. The collected gas samples were input into the photoacoustic-fluid isolation field control module at a flow rate of 0.25 L / min via a constant current sampling device. The isolation field control chamber of this module maintained the sample gas temperature at 25±0.3℃ and humidity at 40±3% through a voltage regulator and micro-circulation airflow channel, thereby reducing the impact of external airflow and temperature / humidity fluctuations on detection sensitivity. After steady-state pretreatment, the SF6 gas sample was introduced into a self-tuning detection unit for photoacoustic signal detection.

[0115] The self-tuning detection unit comprises a main detection cavity, a buffer cavity, and a double-sided composite photothermal response film disposed between the two cavities. The main detection cavity has a volume of 120 mL, the buffer cavity has a volume of 50 mL, and the film thickness is 25 μm. Driven by a photoacoustic signal, the film can achieve a thermo-acoustic coupling response, automatically adjusting the resonant frequency of the main detection cavity to maintain signal stability. During testing, under ambient temperature fluctuations of ±2℃, the resonant frequency drift was less than 0.08%, an improvement of approximately 30% compared to traditional fixed-cavity photoacoustic detection structures.

[0116] After differential detection by dual microphones in the signal acquisition module, the photoacoustic signal undergoes lock-in amplification and digital filtering, reducing the noise amplitude to less than 2.5% of the original signal. Subsequently, the data acquisition and synchronization module performs timestamp calibration and anomaly removal on the signal, generating a standardized detection dataset containing sampling point locations, SF6 concentrations, temperature, humidity, and air pressure information. This dataset is input into a physically interpretable Transformer leakage inversion model. The model extracts diffusion features through a physically constrained coding layer, correlates spatial and temporal information through a multi-scale attention fusion layer, and finally outputs the leakage source location, leakage intensity, and diffusion confidence level through an interpretable inversion decoding layer.

[0117] At the 65th minute of operation monitoring, the system detected that the SF6 concentration at the circuit breaker gas chamber sampling point increased from 0.00010% to 0.00037%. Model inversion calculations determined that the leak source was located in the bottom sealing gasket area of ​​the circuit breaker, with a leakage intensity of 0.0031 L / min. The system automatically triggered an audible and visual alarm after the concentration exceeded the preset threshold of 0.0003%, with an alarm response time of 1.4 seconds. The alarm information was uploaded to the cloud monitoring platform via the local edge node. The platform recorded the event number, leak source location, and detection data, and pushed the data to the station control center. Subsequent on-site maintenance personnel found that a small leak was indeed caused by aging sealant, with a location deviation of less than 18 cm from the system's prediction, and a leakage intensity error of approximately 6%.

[0118] Table 1. Experimental Data Recording Table for SF6 Gas Leakage Detection at 110kV Switchgear Station

[0119]

[0120] As can be seen from the data in Table 1, the SF6 gas leak detection system based on photoacoustic spectroscopy of this invention exhibits high stability and sensitivity in actual operating environments. During the initial monitoring phase (0 to 30 minutes), the SF6 concentration at each sampling point remained between 0.00010% and 0.00012%, with stable temperature, humidity, and pressure parameters. The detection signal amplitude was approximately 2.3–2.4 mV, indicating that the system was in a normal state with a confidence level consistently above 0.95. This demonstrates that the sampling and steady-state control modules operated smoothly and were unaffected by airflow or temperature disturbances.

[0121] At the 60-minute mark, the SF6 concentration at sampling point SP-03 rapidly increased from the background value to 0.00037%, and the signal amplitude significantly increased to 3.18 mV. The model calculated a leakage intensity of 0.0031 L / min. The system's detection result immediately triggered an alarm, which was continuously confirmed at the 65-minute mark. The concentration further increased to 0.00039%, and the predicted leakage intensity reached 0.0032 L / min, with the confidence level maintained between 0.93 and 0.94. This indicates that the system not only has high detection sensitivity but also maintains stable signal judgment capability. The response time for this leakage event was 1.4 seconds, and the alarm status was timely and accurate, showing good agreement with the results of manual verification.

[0122] After the alarm event was handled, the system continued monitoring for 180 minutes and the entire day (720 minutes). The concentration at point SP-03 recovered to 0.00012%, while the concentrations at other sampling points remained within the background range of 0.00010% to 0.00015%. The signal amplitude remained stable between 2.3 and 2.4 mV, and fluctuations in ambient temperature and humidity had almost no impact on the detection signal. The confidence level gradually increased to 0.99, indicating that the system recovered stable detection after the leak source was repaired. Overall, the data shows that the system of this invention can maintain a high signal-to-noise ratio and low drift characteristics for a long time in multi-point synchronous detection. It has the ability to quickly, reliably, and repeatedly identify trace leaks, and the alarm judgment is accurate. It can realize real-time visual monitoring and intelligent early warning of SF6 leaks in power equipment.

[0123] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for detecting SF6 gas leaks based on a photoacoustic spectroscopy analyzer, characterized in that, include: Intelligent sensor sampling terminals are deployed in the operating environment of power equipment to collect SF6 gas samples in the detected area. The collected SF6 gas samples are input into the photoacoustic-fluid isolation field control unit through a constant current sampling device for steady-state preprocessing to obtain steady-state SF6 gas samples. A steady-state SF6 gas sample is introduced into the self-tuning detection unit of the photoacoustic spectrometer. The self-tuning detection unit includes a main detection cavity, a buffer cavity, and a double-sided composite photothermal response film layer set between the two cavities, and outputs a steady-state amplified SF6 photoacoustic signal. The SF6 photoacoustic signal is subjected to dual-microphone differential detection through the signal acquisition module of the photoacoustic spectrum analyzer. After lock-in amplification, digital filtering and power normalization, a stable SF6 detection signal with high signal-to-noise ratio is obtained. In the data acquisition and synchronization module, the stable SF6 detection signal is time-stamped, abnormal data is removed, and signal consistency is processed to generate a standardized detection dataset containing the spatial location of each sampling point, SF6 concentration value, and environmental parameters. The standardized SF6 detection dataset is input into the leakage source inversion model based on a physically interpretable Transformer structure. The leakage source inversion model includes a physically constrained coding layer, a multi-scale attention fusion layer, and an interpretable inversion decoding layer to form the SF6 leakage source inversion result. The SF6 leakage source inversion results are displayed on the visual monitoring interface. When the SF6 concentration value in the detection results exceeds the preset leakage concentration threshold, an alarm is triggered, an audible and visual alarm signal is output, and the SF6 leakage detection results and alarm information are uploaded to the cloud monitoring platform.

2. The SF6 gas leak detection method based on photoacoustic spectroscopy according to claim 1, characterized in that, The deployment of intelligent sensing sampling terminals in the power equipment operating environment refers to setting up several intelligent sensing sampling terminals in the closed or semi-closed spaces where SF6 gas exists, such as substations, circuit breaker bays, gas-insulated busbars, or cable corridors, according to the airflow direction, equipment layout, and leakage path. The intelligent sensing sampling terminals have built-in temperature, humidity, and airflow sensors and are connected to a constant current sampling device to synchronously collect SF6 gas samples and environmental parameters at different locations.

3. The SF6 gas leak detection method based on photoacoustic spectroscopy according to claim 1, characterized in that, The photoacoustic-fluid isolation field control unit is the front-end gas path module of the photoacoustic spectrometer. It includes an airflow isolation ring structure, a variable pressure field control board, and a micro-circulation channel. The airflow isolation ring structure reduces airflow disturbance, the variable pressure field control board maintains a constant detection gas path pressure, and the micro-circulation channel keeps the temperature, humidity, and density of SF6 gas stable, thereby obtaining a steady-state SF6 gas sample.

4. The SF6 gas leak detection method based on photoacoustic spectroscopy according to claim 1, characterized in that, The output steady-state amplified SF6 photoacoustic signal includes: A steady-state SF6 gas sample is introduced into the main detection chamber and buffer chamber of the self-tuning detection unit. The buffer chamber is equipped with a partitioned pressure equalization chamber and a micro-grid impedance layer to perform pressure tuning and flow rate shaping on the SF6 gas entering the main detection chamber. A mid-infrared modulated laser beam is introduced into the main detection cavity to form a fixed optical path. An annular variable optical path reflector is set on the inner wall of the main detection cavity to limit the round-trip path length of the beam. Temperature and pressure sensors are set on the outer wall of the main detection cavity to collect cavity state parameters. A double-sided composite photothermal response film is set between the main detection cavity and the buffer cavity. One side of the double-sided composite photothermal response film faces the main detection cavity and is integrated with the annular micro-thermal driving electrode and temperature sensing unit. The other side faces the buffer cavity and is integrated with the micro-piezoelectric actuator. The surface reflection characteristics of the film are changed by the linkage between the temperature sensing unit and the micro-thermal driving electrode, and the spatial position of the film is finely adjusted by the micro-piezoelectric actuator. During the linkage adjustment process, based on the changes in amplitude and phase of the acoustic signal collected in the main detection cavity and the changes in the state parameters of the main detection cavity and the buffer cavity, the ring variable optical path reflector, the micro-thermal driving electrode and the micro piezoelectric actuator are controlled to work together to output a stable SF6 photoacoustic raw signal. The original SF6 photoacoustic signal is amplified and stabilized once in the self-tuning detection unit, and then packaged and output together with the state parameters of the main detection cavity and the buffer cavity to form a steady-state amplified SF6 photoacoustic signal.

5. The SF6 gas leak detection method based on photoacoustic spectroscopy according to claim 1, characterized in that, The process of obtaining a stable SF6 detection signal with a high signal-to-noise ratio includes: The SF6 photoacoustic raw signal is applied to a dual-microphone differential array set on both sides of the main detection cavity. The dual microphones are symmetrically installed with a half-wavelength spacing corresponding to the target acoustic resonance frequency, and are independently amplified by a phase-synchronized low-noise preamplifier driven by the same clock source. Conductive protective rings and shielding layers are set around the dual microphones, and a miniature Helmholtz blind cavity is set between the dual microphones to attenuate flow noise. Primary cancellation of common-mode noise is achieved through hardware differential bridging with reverse-phase wiring. The reference signal is obtained from the modulated laser drive end of the self-tuning detection unit. The phase-locked detection is performed on the electrical signal after differential bridging of the dual microphones to obtain the in-phase component and the quadrature component. Bandpass digital filtering and anti-aliasing processing are performed at a fixed sampling frequency. The filtering center is set to the frequency consistent with the acoustic resonance of the main detection cavity. Adaptive weighting synthesis is performed on the filtered in-phase and quadrature components to form a differential amplitude signal. The differential amplitude signal is then normalized by the optical power monitoring channel built into the photoacoustic spectroscopy analyzer. The gain and reference phase of the preamplifier are automatically adjusted according to real-time environmental parameters. Narrow-band calibration excitation is injected from the micro piezoelectric actuator in the buffer cavity at fixed time intervals. The dual-channel amplitude and phase registration coefficients are updated using the narrow-band calibration excitation without interfering with the measurement bandwidth. The normalized differential amplitude, phase, timestamp, and corresponding sampling end identifier are used as a stable SF6 detection signal with high signal-to-noise ratio.

6. The SF6 gas leak detection method based on photoacoustic spectroscopy according to claim 1, characterized in that, The generation of the standardized detection dataset, which includes the spatial location of each sampling point, SF6 concentration values, and environmental parameters, includes: It receives stable SF6 detection signals and corresponding timestamps, sampling end identifiers, status parameters of the main detection cavity and buffer cavity, temperature parameters, humidity parameters and airflow parameters, establishes a unified time axis based on the central control clock of the photoacoustic spectrometer, and compensates for the transmission delay and sampling jitter of each channel to generate the original record after time alignment. The original records are subjected to anomaly data removal, including removing data segments corresponding to missing timestamps, abrupt amplitude changes, phase jumps, power monitoring mismatches, and sensor offline conditions. Short-term missing segments are filled by interpolation with adjacent valid segments to obtain valid records after anomaly removal. Based on the identifiers of each sampling terminal and the installation coordinates in the power equipment operating environment, the effective records are mapped into spatial records, the sampling step size is unified, cross-channel amplitude and phase consistency correction and channel gain consistency calibration are completed, and the spatial records after consistency processing are output. The spatial records are associated with temperature, humidity and airflow parameters from intelligent sensor sampling terminals. Standard entries are generated according to the sampling terminal number, and the units of measurement and range are standardized for each field to form a set of compliant entries. The set of compliant items is cataloged and stored in chronological order and sampling order to generate a standardized test dataset containing the spatial location of each sampling point, SF6 concentration-related measurements, and environmental parameters.

7. The SF6 gas leak detection method based on photoacoustic spectroscopy according to claim 1, characterized in that, The inversion results for the formation of SF6 leakage sources include: Receive standardized detection dataset, establish a spatiotemporal sequence according to sampling time order and sampling end spatial coordinates, extract SF6 concentration, airflow velocity direction, temperature, humidity and air pressure of each sampling point, and generate input item set; In the physical constraint coding layer, the input entry set is coded in two levels. The first level of coding generates topological features based on the spatial topological relationship of the sampling points and the connectivity of the airflow channels. The second level of coding generates temporal features based on the joint changes of concentration, temperature and air pressure. The topological features and temporal features are constrained and fused based on the criterion of leakage accessibility to form a spatiotemporal feature sequence. In the multi-scale attention fusion layer, a hierarchical attention structure of local diffusion sub-window and global propagation window is established. Features at different spatial locations within the same time period are fused horizontally, and features at the same spatial location at different times are fused vertically. The stability of environmental parameters is used as a weight to adaptively match the outputs of horizontal and vertical fusion to obtain feature representations. In the interpretable inversion decoding layer, a candidate leak source grid is constructed and constrained by the device geometry boundary and airflow direction. The feature representation is then sparsified and inverted to output the three-dimensional coordinates of the leak source, the leak intensity estimate, and the diffusion path confidence. At the same time, interpretable visualization results containing diffusion direction, diffusion coverage, and boundary consistency are generated. The interpretable visualization results are checked for consistency. The check includes consistency of multi-point concentration back-substitution, consistency of diffusion path and airflow direction, and consistency of equipment geometric boundary matching. If the check passes, the leakage source inversion result is output. If the check fails, the candidate leakage source mesh is automatically updated based on the feature representation and the inversion and check process is repeated until the result meets the consistency requirements.

8. The SF6 gas leak detection method based on photoacoustic spectroscopy according to claim 1, characterized in that, When the SF6 concentration value in the detection result exceeds the preset leakage concentration threshold, an alarm is triggered, and an audible and visual alarm signal is output, including: The visualization monitoring interface displays the SF6 leakage source inversion results, records the display time, sampling end identifier and device location information, and generates event identifiers for local event registration. The real-time detected SF6 concentration value is compared with the preset leakage concentration threshold and graded threshold set. When any threshold condition is met, the alarm module triggers an alarm and outputs an audible and visual alarm signal according to the corresponding threshold level. The alarm level and current confidence level are marked on the monitoring interface. The event identifier, SF6 leakage source inversion results, confidence level, alarm level, concentration value, threshold parameter, time information and sampling end identifier are compiled into an alarm data frame and uploaded to the cloud monitoring platform through the communication link. After receiving the platform's confirmation, the data is archived locally.

9. An SF6 gas leak detection system based on a photoacoustic spectroscopy analyzer, comprising the SF6 gas leak detection method based on a photoacoustic spectroscopy analyzer as described in any one of claims 1 to 8, characterized in that, Includes the following modules: The intelligent sampling module is used to deploy multiple sampling points in the operating environment of power equipment to collect SF6 gas samples in real time. The fluid steady-state control module is used to perform steady-state preprocessing on SF6 gas samples and output steady-state SF6 gas samples. The self-tuning detection module is used to import steady-state SF6 gas samples into the photoacoustic spectrometer for detection and output steady-state amplified SF6 photoacoustic signals. The signal differential processing module is used to perform dual-microphone differential detection on the SF6 photoacoustic signal, and obtain a stable SF6 detection signal through phase-locked amplification, digital filtering and power normalization. The data synchronization module is used to perform time calibration, anomaly removal, and consistency processing on stable SF6 detection signals to generate a standardized detection dataset. The leakage inversion and alarm module is used to input standardized data into the leakage source inversion model for inversion analysis and display the results on the monitoring interface. When the concentration exceeds the limit, it triggers an audible and visual alarm and uploads the data to the cloud monitoring platform.

Images