A gas detection method based on photoacoustic spectroscopy principle

Abstract

This invention discloses a gas detection method based on photoacoustic spectroscopy. The method employs a detection device containing relays, which includes an IED processing unit to collect fault gas concentration data, dynamically calculate hourly and daily gas production rates, and run a multi-level fault fusion diagnostic algorithm based on the D-S evidence theory. The algorithm uses the three-ratio feature obtained from the decoupled gas concentration calculation, the dynamic gas production rate trend feature, the oil temperature change rate feature, and the gas collecting chamber pressure change rate feature as four independent evidence bodies. For each evidence body, a basic probability allocation function corresponding to the fault type is constructed. The D-S evidence combination rule is used to fuse and decide on the multi-source evidence, outputting the fault type and confidence level. A preliminary fault determination result is obtained by running the three-ratio method, and then a secondary verification is performed using the David's triangle diagnostic model. Upon detection of C2H2 gas, a high-risk discharge warning is immediately output.

Description

Technical Field

[0001] This invention relates to the field of online monitoring of power transformer status, and in particular to a gas detection method based on the photoacoustic spectroscopy principle. Background Technology

[0002] Oil-immersed power transformers are core and critical equipment in power transmission and distribution networks, and their operational reliability directly determines the safety of power supply. Under long-term energized operation, if internal faults such as overheating, partial discharge, or arc breakdown occur in the transformer's internal insulating oil and paper insulation materials, the insulating medium will decompose, generating characteristic fault gases such as CO2, CO, CH4, C2H6, C2H4, and C2H2. These fault gases continuously migrate upwards, accumulating at the highest point of the transformer's oil conservator connection pipeline, and eventually gathering inside the gas collection chamber of the gas relay.

[0003] Traditional mechanical gas relays rely on a float to lower the oil level and trigger a light gas volume alarm, while a fault-induced oil flow pushes a baffle to trip the circuit breaker for heavy gas protection. They are the standard main protection device for oil-immersed transformers. However, traditional mechanical gas relays have significant drawbacks: they can only output passive switching signals and cannot identify the composition, concentration, and gas generation rate of the accumulated gas; they cannot distinguish between air released from oil temperature changes and fault-specific gases generated by insulation decomposition, leading to frequent false alarms for light gas without actual faults and dulling the sensitivity of fault alarms. Furthermore, after a traditional mechanical gas relay trips, maintenance personnel must shut down the power on-site, manually collect the gas, and send it to a laboratory for gas chromatography analysis to determine the fault type. This results in a significant delay in fault assessment, and latent faults cannot be addressed in advance, easily leading to transformer damage and power grid outages.

[0004] Currently, the mainstream online gas monitoring equipment in the transformer oil industry is an external oil-gas separation monitoring device, which requires additional oil and gas pipelines. The equipment is bulky, the on-site construction and modification are complex, and the modification cost is high. Moreover, it cannot take advantage of the gas relay's structure to efficiently collect free gas. Furthermore, the external monitoring equipment is independent of the original gas relay and the data is not linked, resulting in equipment redundancy and a disconnect between the protection and monitoring systems.

[0005] Therefore, a method or apparatus is needed to solve the above problems. Summary of the Invention

[0006] The present invention addresses the aforementioned shortcomings of existing technologies by designing a method for gas detection based on photoacoustic spectroscopy that can effectively reduce the probability of equipment malfunction and false alarms, provide early warning of latent faults in transformers, and comprehensively improve the operational safety level of oil-immersed transformers.

[0007] The technical solution of the present invention is: a gas detection method based on the photoacoustic spectroscopy principle, which uses a detection device containing a relay to achieve detection. The relay includes a relay housing, and the relay housing is provided with a gas collection chamber, a light gas float assembly and a heavy gas baffle trip assembly. The relay housing is also provided with a photoacoustic spectroscopy gas detection module. The photoacoustic spectroscopy gas detection module includes a resonant photoacoustic cell. An input optical window and an output optical window are respectively provided at both ends of the resonant photoacoustic cell. A laser component and an optical switch are provided in the incident direction of the input optical window. A sound transmission hole is provided on the side wall of the resonant photoacoustic cell. The sound transmission hole is connected to the inner cavity of the resonant photoacoustic cell. A micro-sound sensor is provided at the end of the sound transmission hole. The photoacoustic spectroscopy gas detection module also includes a sampling gas pump. The inlet of the sampling gas pump is connected to the gas collection chamber through a pipeline, and its outlet is connected to the inlet pipeline of the resonant photoacoustic cell through a pipeline. A solenoid valve is also installed on the outlet pipeline of the resonant photoacoustic cell. The laser assembly, optical switch, microphone, sampling gas pump, and solenoid valve are all electrically connected to the IED processing unit, and a temperature and pressure detection assembly is also provided in the gas collection cavity. The detection method is operated according to the following steps: S1: First, the IED processing unit collects the concentration data of the fault gas, as well as the liquid level, temperature and pressure values ​​in the gas collection chamber, and uses the wavelet threshold-LSTM multi-component gas cross-interference decoupling algorithm to perform adaptive wavelet threshold denoising on the original photoacoustic signal collected by the micro-sound sensor, constructs a continuously adjustable improved threshold function, and filters out noise caused by on-site vibration and electromagnetic interference while retaining the signal feature details. Then, the denoised 6-channel wavelength photoacoustic time sequence signal is input into the pre-trained lightweight LSTM neural network. The time sequence feature extraction capability of the lightweight LSTM neural network decouples the cross-interference caused by the overlap of the absorption spectra of each gas, and outputs the corrected true concentration of each component gas. S2: Then, dynamically calculate the hourly and daily gas production rates, and run the DS evidence theory multi-level fault fusion diagnosis algorithm. The three ratios of the decoupled gas concentration, the dynamic gas production rate trend, the oil temperature change rate, and the gas collection chamber pressure change rate are used as four independent evidence bodies. For each evidence body, a basic probability allocation function corresponding to the fault type is constructed. The DS evidence combination rule is used to fuse and decide on the multi-source evidence, and the fault type and confidence level are output. S3: The three-ratio method is used to obtain the preliminary fault judgment result, and then the David triangle diagnostic model is used for secondary verification. S4: When C2H2 gas is detected, a high-risk discharge warning will be issued immediately; when only air accumulation is detected and there is no characteristic fault gas growth, it will be judged as normal temperature rise and gas evolution, and the light gas warning will be ignored. S5: Summarize information on gas concentration, gas production rate, fault diagnosis conclusions, and early warning status, and transmit the above information to the substation monitoring backend through the remote data transmission module built into the IED processing unit.

[0008] The temperature and pressure detection assembly includes a liquid level sensor, a temperature sensor, and a pressure sensor.

[0009] The laser assembly is capable of emitting lasers with characteristic absorption wavelengths matching those of CO2, CO, CH4, C2H6, C2H4, and C2H2.

[0010] Compared with the prior art, the present invention has the following advantages: This gas detection method based on photoacoustic spectroscopy addresses the various problems of traditional gas relays by designing a special relay structure and utilizing this relay in conjunction with specific procedures for gas detection. It incorporates a photoacoustic spectroscopy detection system within the relay, directly collecting the free fault gas floating in the transformer. This eliminates the need for an external oil-gas separator. Therefore, the overall appearance, mounting flange, and piping interfaces of this device are completely identical to traditional gas relays, allowing for direct replacement with minimal modification work. It is suitable for upgrading and retrofitting various existing oil-immersed transformers. At the same time, it completely retains the original mechanical gas main protection circuit of the relay, and the relay protection performance is not affected. This method can process the collected data to accurately distinguish between gas evolution and normal gas evolution, which can significantly reduce the probability of false alarms. In addition, it can monitor characteristic gases such as acetylene in real time, thereby realizing early warning of high-risk faults such as electric arc and partial discharge. Since the core component for detecting faulty gases is the resonant photoacoustic cell, the detection process has many advantages such as resistance to vibration, environmental noise, and strong electromagnetic interference. Combined with temperature and pressure compensation correction in the data processing method, it can still achieve long-term, stable, and high-precision detection even under harsh conditions such as outdoor wide temperature range or complex electromagnetic conditions. In conclusion, this detection method has many advantages and is particularly suitable for widespread application in this field, with a very broad market prospect. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of the structure of an embodiment of the present invention.

[0012] Figure 2 This is a schematic diagram of the structure of the photoacoustic spectroscopy gas detection module in an embodiment of the present invention.

[0013] 1. Relay housing; 2. Gas collection chamber; 3. Light gas float assembly; 4. Heavy gas baffle trip assembly; 5. Photoacoustic spectroscopy gas detection module; 6. Resonant photoacoustic cell; 7. Input light window; 8. Output light window; 9. Laser assembly; 10. Optical switch; 11. Micro-sound sensor; 12. Sampling gas pump; 13. Solenoid valve; 14. IED processing unit; 15. Liquid level sensor; 16. Temperature sensor; 17. Pressure sensor. Detailed Implementation

[0014] Specific embodiments of the present invention will now be described in conjunction with the accompanying drawings. Figure 1 , Figure 2 As shown: A gas detection method based on photoacoustic spectroscopy principle is implemented by using a detection device containing a relay. The relay includes a relay housing 1, and the relay housing 1 is provided with a gas collection chamber 2, a light gas float assembly 3 and a heavy gas baffle trip assembly 4. The relay housing 1 is also provided with a photoacoustic spectroscopy gas detection module 5. The photoacoustic spectroscopy gas detection module 5 includes a resonant photoacoustic cell 6. An input optical window 7 and an output optical window 8 are respectively provided at both ends of the resonant photoacoustic cell 6. A laser assembly 9 and an optical switch 10 are provided in the incident direction of the input optical window 7. A sound transmission hole is provided on the side wall of the resonant photoacoustic cell 6. The sound transmission hole is connected to the inner cavity of the resonant photoacoustic cell 6. A micro-sound sensor 11 is provided at the end of the sound transmission hole. The photoacoustic spectroscopy gas detection module 5 also includes a sampling gas pump 12. The inlet of the sampling gas pump 12 is connected to the gas collection chamber 2 through a pipeline, and its outlet is connected to the inlet pipeline of the resonant photoacoustic cell 6 through a pipeline. The outlet pipeline of the resonant photoacoustic cell 6 is also equipped with a solenoid valve 13. The laser assembly 9, optical switch 10, microphone sensor 11, sampling gas pump 12 and solenoid valve 13 are all electrically connected to the IED processing unit 14, and a temperature and pressure detection assembly is also provided in the gas collection chamber 2. The temperature and pressure detection assembly includes a liquid level sensor 15, a temperature sensor 16, and a pressure sensor 17.

[0015] The laser assembly 9 is capable of emitting lasers with characteristic absorption wavelengths matching those of CO2, CO, CH4, C2H6, C2H4, and C2H2.

[0016] When a fault occurs inside the transformer, the characteristic gas generated by the fault will rise and accumulate in the gas collection chamber 2 inside the relay housing 1. The liquid level sensor 15, temperature sensor 16 and pressure sensor 17 in the temperature and pressure detection assembly installed in the gas collection chamber 2 monitor various parameters in the gas collection chamber 2 in real time. When the IED processing unit 14 determines that gas accumulation has occurred in the gas collection chamber 2 based on the above parameters, it starts sampling and detection operations. First, the IED processing unit 14 controls the sampling gas pump 12 to work. The sampling gas pump 12 pumps the characteristic gas in the gas collection chamber 2 to the inner cavity of the resonant photoacoustic cell 6 to complete the gas sample collection. The program in the IED processing unit 14 automatically runs the temperature adaptive resonant frequency tracking algorithm and dynamically matches the laser modulation frequency to ensure that the resonant photoacoustic cell 6 is always in the optimal resonant working state. Based on the laser modulation frequency obtained in the previous process, the IED processing unit 14 controls the laser component 9 to time-division modulate the emission. After passing through the optical switch 10, the laser is introduced into the inner cavity of the resonant photoacoustic cell 6, which excites the characteristic gas in it to generate a photoacoustic signal. The micro-sound sensor 11 collects the photoacoustic signal, completes conditioning and amplification, and inputs it into the IED processing unit 14. The pre-built program performs adaptive wavelet threshold denoising on the raw photoacoustic signal collected by the micro-sound sensor 11, constructs a continuously adjustable improved threshold function, and filters out noise caused by on-site vibration and electromagnetic interference while preserving the signal feature details. Then, the denoised 6-channel wavelength photoacoustic time-series signal is input into a pre-trained lightweight LSTM neural network. The time-series feature extraction capability of the lightweight LSTM neural network decouples the cross interference caused by the overlap of gas absorption spectral lines and outputs the corrected true concentration of each component gas. The hourly and daily gas production rates are dynamically calculated, and the DS evidence theory multi-level fault fusion diagnosis algorithm is run. The three ratios of the decoupled gas concentration, the dynamic gas production rate trend, the oil temperature change rate, and the gas collection chamber pressure change rate are used as four independent evidence bodies. For each evidence body, a basic probability allocation function corresponding to the fault type is constructed. The DS evidence combination rule is used to fuse and decide on the multi-source evidence, and the fault type and confidence level are output. The initial fault determination result is obtained by the three ratio method, and then a second verification is performed by the satellite triangle model to determine the specific fault type of the transformer. When C2H2 gas is detected, a high-risk discharge warning is immediately issued; when only air accumulation is detected and there is no characteristic fault gas growth, it is determined to be normal temperature rise and gas evolution, and the light gas warning is ignored. When the accumulated gas volume reaches the threshold, the light gas float assembly 3 is triggered, and this relay will issue a light gas alarm. When a serious fault occurs in the transformer and an impact oil flow is generated, the heavy gas damper trip assembly 4 will be activated, and this relay will output a trip protection signal to perform a trip action.

[0017] Finally, the information on gas concentration, gas production rate, fault diagnosis conclusions, and early warning status is summarized and transmitted to the substation monitoring backend through the remote data transmission module built into the IED processing unit 14.

Claims

1. A gas detection method based on the photoacoustic spectroscopy principle, employing a detection device containing a relay for detection, characterized in that: The relay includes a relay housing (1), and a gas collection chamber (2), a light gas float assembly (3), and a heavy gas baffle trip assembly (4) are provided inside the relay housing (1). A photoacoustic spectroscopy gas detection module (5) is also provided inside the relay housing (1). The photoacoustic spectroscopy gas detection module (5) includes a resonant photoacoustic cell (6). An input optical window (7) and an output optical window (8) are respectively provided at both ends of the resonant photoacoustic cell (6). A laser assembly (9) and an optical switch (10) are provided in the incident direction of the input optical window (7). A sound transmission hole is provided on the side wall of the resonant photoacoustic cell (6). The sound transmission hole is connected to the inner cavity of the resonant photoacoustic cell (6). A micro-sound sensor (11) is provided at the end of the sound transmission hole. The photoacoustic spectroscopy gas detection module (5) also includes a sampling gas pump (12). The inlet of the sampling gas pump (12) is connected to the gas collection chamber (2) through a pipeline, and its outlet is connected to the inlet pipeline of the resonant photoacoustic cell (6) through a pipeline. A solenoid valve (13) is also provided on the outlet pipeline of the resonant photoacoustic cell (6). The laser assembly (9), optical switch (10), micro-sound sensor (11), sampling gas pump (12) and solenoid valve (13) are all electrically connected to the IED processing unit (14), and a temperature and pressure detection assembly is also provided in the gas collection chamber (2); The detection method is operated according to the following steps: S1: First, the IED processing unit (14) collects the concentration data of the fault gas, as well as the liquid level, temperature and pressure values ​​in the gas collection chamber (2), and uses the wavelet threshold-LSTM multi-component gas cross-interference decoupling algorithm to perform adaptive wavelet threshold denoising on the original photoacoustic signal collected by the micro-sound sensor (11), constructs a continuously adjustable improved threshold function, and filters out the noise caused by field vibration and electromagnetic interference while retaining the signal feature details. Then, the denoised 6-channel wavelength photoacoustic time sequence signal is input into the pre-trained lightweight LSTM neural network. The cross-interference caused by the overlap of the absorption spectra of each gas is decoupled through the time sequence feature extraction capability of the lightweight LSTM neural network, and the corrected true concentration of each component gas is output. S2: Then, dynamically calculate the hourly and daily gas production rates, and run the DS evidence theory multi-level fault fusion diagnosis algorithm. The three ratios of the decoupled gas concentration, the dynamic gas production rate trend, the oil temperature change rate, and the gas collection chamber pressure change rate are used as four independent evidence bodies. For each evidence body, a basic probability allocation function corresponding to the fault type is constructed. The DS evidence combination rule is used to fuse and decide on the multi-source evidence, and the fault type and confidence level are output. S3: The three-ratio method is used to obtain the preliminary fault judgment result, and then the David triangle diagnostic model is used for secondary verification. S4: When C2H2 gas is detected, a high-risk discharge warning will be issued immediately; When only air accumulation is detected and there is no characteristic fault gas growth, it is determined to be normal temperature rise gas evolution, and the light gas warning is ignored. S5: Summarize the information on gas concentration, gas production rate, fault diagnosis conclusion and early warning status, and transmit the above information to the substation monitoring backend through the remote data transmission module built into the IED processing unit (14).

2. The gas detection method based on photoacoustic spectroscopy according to claim 1, characterized in that: The temperature and pressure detection assembly includes a liquid level sensor (15), a temperature sensor (16), and a pressure sensor (17).

3. The gas detection method based on photoacoustic spectroscopy according to claim 2, characterized in that: The laser assembly (9) is capable of emitting lasers with characteristic absorption wavelengths matching those of CO2, CO, CH4, C2H6, C2H4 and C2H2, respectively.

Images