Transformer oil dissolved gas surface enhanced Raman spectrum in-situ monitoring method and system

Abstract

The invention belongs to the technical field of power equipment state monitoring, and discloses a transformer oil dissolved gas surface enhanced Raman spectrum in-situ monitoring method and a transformer oil dissolved gas surface enhanced Raman spectrum in-situ monitoring system. The system comprises a membrane degassing unit, a Raman sensing unit, a light path unit, a data processing unit and a calibration early warning unit, the method comprises the following steps: firstly preparing a gold nanostar array sensing substrate with bifunctional groups to realize selective adsorption and signal enhancement of characteristic gas, then completing in-situ oil-gas separation through the membrane degassing unit, collecting a Raman spectrum through the light path unit, and calculating the Raman spectrum of the gold nanostar array sensing substrate according to the Raman spectrum. Spectral analysis and quantification are completed based on multiple algorithms, and fault early warning is achieved in combination with temperature and pressure compensation and an industrial standard threshold value. According to the scheme, the problems of lagging, weak interference resistance and the like of traditional monitoring are solved, high sensitivity and high specificity are achieved, full-process automatic closed-loop monitoring can be achieved, and accurate data support is provided for operation and maintenance of the transformer.

Description

Technical Field

[0001] This invention relates to the field of power equipment condition monitoring and Raman spectroscopy sensing technology, specifically to a method and system for in-situ monitoring of dissolved gases in transformer oil using surface-enhanced Raman spectroscopy, which is suitable for real-time early warning of latent faults inside oil-immersed transformers and quantitative analysis of multiple characteristic gases. Background Technology

[0002] Power transformers are core equipment in power systems, and their operating status directly affects the safety and stability of the power grid. Internal faults in transformers can lead to the cracking of insulating oil, producing characteristic dissolved gases such as H2, CH4, and C2H2. By monitoring the composition and concentration of these gases, early fault diagnosis can be achieved.

[0003] Traditional monitoring technologies often employ offline sampling followed by gas chromatography analysis. This method has significant limitations: firstly, the sampling process is time-consuming, resulting in monitoring lag and making it difficult to provide real-time early warning of faults; secondly, the sampling and transportation process is prone to introducing external interference, affecting the accuracy of the test results.

[0004] Some online monitoring technologies are also applied in the field, but existing online technologies have many shortcomings. For example, the sensing elements are not selective for characteristic gases and are easily interfered with by impurities or non-target gases in the oil; some technologies have insufficient signal detection sensitivity, making it difficult to identify low-concentration fault gases; and some technologies do not consider the influence of temperature and pressure environment during transformer operation, resulting in large deviations in quantitative results and failing to provide accurate data support for fault diagnosis. Therefore, there is an urgent need for a transformer oil dissolved gas monitoring technology that is in-situ, real-time, highly sensitive, and highly selective.

[0005] To address this, a method and system for in-situ monitoring of dissolved gases in transformer oil using surface-enhanced Raman spectroscopy are proposed. Summary of the Invention

[0006] The present invention aims to solve the problems mentioned in the background art by providing a method and system for in-situ monitoring of dissolved gases in transformer oil using surface-enhanced Raman spectroscopy.

[0007] The specific technical solution is as follows: A method for in-situ monitoring of dissolved gases in transformer oil using surface-enhanced Raman spectroscopy includes the following steps: Step 1: Prepare a multi-component selective surface-enhanced Raman spectroscopy sensing substrate, grow a gold nanostar array on the silicon-based array surface, and sequentially graft thiol functional groups and amino functional groups to form selective adsorption sites for characteristic gases in transformer oil. Step 2: The monitoring probe integrating the surface-enhanced Raman spectroscopy sensing substrate is pre-embedded in the fault-prone area inside the transformer oil tank. The dissolved gas in the transformer oil is separated in situ through the membrane degassing unit, so that the characteristic gas permeates to the surface of the surface-enhanced Raman spectroscopy sensing substrate. Step 3: Emit an excitation laser to the surface-enhanced Raman spectroscopy sensing substrate through the optical path unit to collect the surface-enhanced Raman scattering spectrum of the characteristic gas; Step 4: Preprocess the acquired raw spectra, separate overlapping feature peaks using a deconvolution algorithm, and then use partial least squares discriminant analysis combined with support vector machine algorithm to achieve qualitative identification and quantitative analysis of multi-feature gases; Step 5: Perform environmental compensation on the quantitative results through the temperature-pressure calibration module, output gas component and concentration data, and complete fault warning.

[0008] In the above-mentioned in-situ monitoring method of surface-enhanced Raman spectroscopy for dissolved gases in transformer oil, the gold nanostar array in step 1 has a particle size of 120nm-180nm and a tip curvature radius of 8nm-12nm, and the silicon-based array has a period of 1.8μm-2.2μm.

[0009] In the above-mentioned in-situ monitoring method of transformer oil dissolved gas by surface-enhanced Raman spectroscopy, the membrane component of the membrane degassing unit in step 2 is a polytetrafluoroethylene hollow fiber membrane with a pore size of 0.15μm-0.25μm and an oil-gas separation efficiency of not less than 85%.

[0010] In the above-described in-situ monitoring method for dissolved gases in transformer oil using surface-enhanced Raman spectroscopy, the excitation laser in step 3 is a 785nm near-infrared laser with an output power of 40mW-60mW, and the detection range of the surface-enhanced Raman scattering spectrum is 200cm. -1 -4000cm -1 .

[0011] The present invention also provides an in-situ monitoring system for surface-enhanced Raman spectroscopy of dissolved gases in transformer oil, for implementing the above-mentioned in-situ monitoring method for surface-enhanced Raman spectroscopy of dissolved gases in transformer oil, including a membrane degassing unit, a surface-enhanced Raman spectroscopy sensing unit, an optical path unit, a data processing unit, and a calibration and early warning unit; The oil inlet of the membrane degassing unit is connected to the inside of the transformer oil tank, and the gas outlet is sealed and connected to the surface-enhanced Raman spectroscopy sensing unit. The surface-enhanced Raman spectroscopy sensing unit includes the multi-component selective surface-enhanced Raman spectroscopy sensing substrate and the sealed sensing cavity as described in claim 1, wherein the surface-enhanced Raman spectroscopy sensing substrate is fixed inside the sealed sensing cavity; The optical path unit includes a laser, an optical fiber coupler, an optical fiber probe, and a spectral detector. The output end of the laser is connected to the optical fiber probe through the optical fiber coupler. The optical fiber probe faces the surface-enhanced Raman spectroscopy sensing substrate. The spectral detector is connected to the sealed sensing cavity through a Raman signal collecting optical fiber. The data processing unit has a built-in spectral preprocessing module, a peak deconvolution module, and a multi-model fusion quantitative module. The calibration and early warning unit includes a temperature sensor, a pressure sensor, and a fault threshold database.

[0012] In the above-mentioned transformer oil dissolved gas surface-enhanced Raman spectroscopy in-situ monitoring system, the laser is a narrow linewidth semiconductor laser with a linewidth not exceeding 0.001 pm, and the spectral detector is a charge-coupled device with a pixel count not less than 2000×256.

[0013] The above-mentioned transformer oil dissolved gas surface-enhanced Raman spectroscopy in-situ monitoring system includes a data processing unit whose spectral preprocessing module comprises a baseline correction submodule and a smoothing and denoising submodule, wherein the smoothing and denoising submodule employs the Savitzky-Golay filtering algorithm.

[0014] The above-mentioned transformer oil dissolved gas surface-enhanced Raman spectroscopy in-situ monitoring system includes a temperature sensor in the calibration and early warning unit with a measurement range of -25℃ to 70℃ and a pressure sensor with a measurement range of 0.05MPa to 0.35MPa. The fault threshold database is built according to the DL / T722-2018 standard.

[0015] In the above-mentioned transformer oil dissolved gas surface-enhanced Raman spectroscopy in-situ monitoring system, the inner wall of the sealed sensing cavity is coated with an anti-reflective film, the reflectivity of the anti-reflective film does not exceed 5%, and the overall volume of the sensing cavity does not exceed 5 cm³.

[0016] In the above-mentioned transformer oil dissolved gas surface-enhanced Raman spectroscopy in-situ monitoring system, the oil inlet end of the membrane degassing unit is provided with a microfiltration component, the pore size of which is 1μm-2μm, for intercepting solid impurities in the transformer oil.

[0017] The present invention has the following beneficial effects: 1. Solving the problem of lag in traditional monitoring: By pre-embedded probes combined with the in-situ oil-gas separation design of the membrane degassing unit, online in-situ monitoring of dissolved gases in transformer oil is realized. There is no need for offline sampling and testing. Gas composition data can be obtained in real time, which saves time for the early detection of internal transformer faults and avoids the expansion of faults due to monitoring lag.

[0018] 2. Enhanced monitoring sensitivity and specificity: The structural design of the gold nanostar array enhances the Raman signal, enabling even low-concentration fault characteristic gases to generate identifiable Raman signals; the grafting of bifunctional groups endows the sensing substrate with selective adsorption capabilities for target gases, effectively filtering out interference from non-target components and impurities in the oil, thus significantly improving the monitoring sensitivity and specificity.

[0019] 3. Ensuring the accuracy of detection data: The combination of a narrow-linewidth laser and a high-pixel spectral detector ensures the stability and accuracy of Raman signal acquisition; multi-step spectral preprocessing eliminates interference such as baseline drift and signal noise; peak deconvolution technology solves the problem of overlapping characteristic peaks; and multi-algorithm fusion analysis improves the reliability of qualitative and quantitative analysis of multi-component gases; at the same time, the temperature and pressure environment compensation mechanism eliminates the impact of transformer operating environment fluctuations on detection results, further ensuring the accuracy of the data.

[0020] 4. Enhanced system practicality and durability: The microfiltration component at the oil inlet can intercept solid impurities, preventing membrane module clogging or damage, extending the service life of the membrane degassing unit, and reducing the frequency of later system maintenance; the anti-reflective membrane design of the sealed sensing cavity ensures the purity of signal acquisition, while the miniaturized cavity structure facilitates placement in the narrow, fault-prone areas of the transformer tank, improving the system's field adaptability; the fault threshold database built according to industry standards provides professional and authoritative judgment criteria for fault early warning, enhancing the engineering practicality of the technical solution.

[0021] 5. Forming an integrated monitoring closed loop: From in-situ oil and gas separation and signal enhancement acquisition to accurate data analysis, environmental compensation, and fault early warning, the entire technology forms a complete monitoring closed loop, which can realize the fully automated processing from gas detection to fault diagnosis. It provides integrated technical support for transformer operation and maintenance, reduces the workload of operation and maintenance personnel, and improves the operation and maintenance management level of power grid equipment. Attached Figure Description

[0022] Figure 1 A flowchart of the in-situ monitoring method for dissolved gases in transformer oil using surface-enhanced Raman spectroscopy provided in this embodiment of the invention; Figure 2 The graph shows the sensitivity improvement. Figure 3 A bar chart showing the selective adsorption effect; Figure 4 This is a graph showing the oil-gas separation efficiency. Figure 5 This is a comparison chart of the environmental compensation effects. Detailed Implementation

[0023] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0024] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual images. They should not be construed as limiting the scope of this application. To better illustrate the embodiments of the present invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual dimensions of the product. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0025] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "inner," and "outer" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present application. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0026] In the description of this invention, unless otherwise explicitly specified and limited, the term "connection" or similar designation indicating a connection between components should be interpreted broadly. For example, it can refer to a fixed connection, a detachable connection, or an integral part; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can refer to the internal communication between two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0027] Example 1 Reference Figures 1-5 ,in, Figure 2 The linear relationship between gas concentration (ppm) and Raman signal intensity (au) is demonstrated to verify the enhanced detection capability of gold nanostar array for low-concentration fault gases. Figure 3 By comparing the differences in adsorption amounts of gases such as H2, CH4, and C2H2 on the bifunctional grafted sensing substrate, the selective adsorption characteristics of the material for specific gases are demonstrated. Figure 4 The correlation between membrane pore size (0.15-0.25 μm) and separation efficiency (%) was revealed, and the design parameters of the membrane degassing unit were optimized. Figure 5 The difference in gas concentration before and after compensation is shown by measuring temperature (10-50℃), which verifies the effectiveness of the temperature and pressure compensation mechanism.

[0028] The in-situ monitoring method for dissolved gases in transformer oil using surface-enhanced Raman spectroscopy provided in this embodiment includes the following steps: Step 1: Prepare a multi-component selective surface-enhanced Raman spectroscopy sensing substrate, grow a gold nanostar array on the silicon-based array surface, and sequentially graft thiol functional groups and amino functional groups to form selective adsorption sites for characteristic gases in transformer oil. Step 2: The monitoring probe of the integrated surface-enhanced Raman spectroscopy sensing substrate is pre-embedded in the fault-prone area inside the transformer oil tank. The dissolved gas in the transformer oil is separated in situ through the membrane degassing unit, so that the characteristic gas can penetrate to the surface of the surface-enhanced Raman spectroscopy sensing substrate. Step 3: Emit an excitation laser to the surface-enhanced Raman spectroscopy sensing substrate through the optical path unit to collect the surface-enhanced Raman scattering spectrum of the characteristic gas; Step 4: Preprocess the acquired raw spectra, separate overlapping feature peaks using a deconvolution algorithm, and then use partial least squares discriminant analysis combined with support vector machine algorithm to achieve qualitative identification and quantitative analysis of multi-feature gases; Step 5: Perform environmental compensation on the quantitative results through the temperature-pressure calibration module, output gas component and concentration data, and complete fault warning.

[0029] By fabricating a bifunctional gold nanostar sensing substrate on a silicon-based array, selective adsorption of characteristic gases in transformer oil can be achieved. At the same time, the Raman signal is enhanced by the surface plasmon effect of gold nanostars. The pre-embedded probe combined with the membrane degassing unit achieves in-situ separation of oil and gas, solving the problem of lag in traditional monitoring that requires sampling. After spectral preprocessing, multi-algorithm joint analysis and temperature and pressure environment compensation, accurate qualitative and quantitative analysis of multiple characteristic gases can be achieved, ultimately enabling early warning of transformer faults and realizing full-process in-situ monitoring of dissolved gases in transformer oil.

[0030] Specifically, in this embodiment, the gold nanostar array in step 1 has a particle size of 120nm-180nm and a tip curvature radius of 8nm-12nm, while the silicon-based array has a period of 1.8μm-2.2μm. By limiting the microstructure of the gold nanostar array and the arrangement parameters of the silicon-based array, the distribution of localized plasmonic "hot spots" on the surface of the gold nanostars can be optimized, enhancing the Raman signal enhancement effect. Simultaneously, the distribution of functional group adsorption sites becomes more regular, thereby improving the sensitivity and signal stability of the sensing substrate for characteristic gas detection.

[0031] Specifically, in this embodiment, the membrane module of the membrane degassing unit in step 2 is a polytetrafluoroethylene hollow fiber membrane with a pore size of 0.15μm-0.25μm and an oil-gas separation efficiency of not less than 85%. Using a polytetrafluoroethylene hollow fiber membrane as the membrane degassing module leverages its excellent chemical stability and oleophobic / permeable properties to efficiently achieve oil-gas separation. This not only prevents the oil phase from entering the detection stage to eliminate interference but also ensures that the characteristic gas can smoothly permeate to the sensing substrate, providing a pure gas sample for subsequent Raman detection and improving the reliability of the monitoring results.

[0032] Specifically, in this embodiment, the excitation laser in step 3 is a 785nm near-infrared laser with an output power of 40mW-60mW, and the detection range of the surface-enhanced Raman scattering spectrum is 200cm. -1 -4000cm -1 By selecting a near-infrared excitation laser in a specific wavelength band, the fluorescence background interference during the detection process can be significantly reduced. At the same time, the matched laser power can ensure the Raman signal intensity while avoiding damage to the sensing substrate and gas components. The corresponding spectral detection range can cover the Raman characteristic peaks of various characteristic gases in transformer oil, achieving comprehensive detection of multi-component gases.

[0033] Example 2 This embodiment provides an in-situ monitoring system for surface-enhanced Raman spectroscopy of dissolved gases in transformer oil, used to implement the in-situ monitoring method for surface-enhanced Raman spectroscopy of dissolved gases in transformer oil in Embodiment 1. The system includes a membrane degassing unit, a surface-enhanced Raman spectroscopy sensing unit, an optical path unit, a data processing unit, and a calibration and early warning unit. The oil inlet of the membrane degassing unit is connected to the inside of the transformer oil tank, and the gas outlet is sealed and connected to the surface-enhanced Raman spectroscopy sensing unit. The surface-enhanced Raman spectroscopy sensing unit includes a multi-component selective surface-enhanced Raman spectroscopy sensing substrate and a sealed sensing cavity as claimed in claim 1, wherein the surface-enhanced Raman spectroscopy sensing substrate is fixed inside the sealed sensing cavity; The optical path unit includes a laser, an optical fiber coupler, an optical fiber probe, and a spectrometer. The output of the laser is connected to the optical fiber probe through the optical fiber coupler. The optical fiber probe faces the surface-enhanced Raman spectroscopy sensing substrate. The spectrometer is connected to the sealed sensing cavity through a Raman signal collecting optical fiber. The data processing unit has a built-in spectral preprocessing module, peak deconvolution module, and multi-model fusion quantitative module; The calibration and early warning unit includes a temperature sensor, a pressure sensor, and a fault threshold database.

[0034] The constructed multi-unit collaborative monitoring system can fully adapt to the entire process requirements of the aforementioned monitoring methods: the membrane degassing unit realizes efficient in-situ separation of oil and gas, the sensing unit provides stable selective detection sites, the optical path unit ensures precise laser excitation and efficient acquisition of Raman signals, the data processing unit completes accurate spectral analysis and quantification, and the calibration and early warning unit eliminates environmental interference and realizes fault judgment, forming an integrated in-situ monitoring capability.

[0035] Specifically, in this embodiment, the connection relationships between the units are as follows: Membrane degassing unit and surface-enhanced Raman spectroscopy sensing unit The outlet of the membrane degassing unit is connected to the air inlet of the sealed sensing cavity of the surface-enhanced Raman spectroscopy sensing unit through a stainless steel sealing flange. An oil-resistant fluororubber sealing ring is installed on the flange sealing surface to ensure the airtightness of the connection and prevent leakage of characteristic gases. The oil inlet of the membrane degassing unit is connected to the pre-set oil outlet inside the transformer tank through a pressure-resistant oil pipe. The oil pipe and the oil outlet are fixed by a double method of welding and compression fitting, which ensures the connection strength and avoids oil leakage.

[0036] Surface-enhanced Raman spectroscopy sensing unit and optical path unit The fiber optic probe and Raman signal collection fiber of the optical path unit are inserted into the sealed sensing cavity through an airtight fiber optic connector, and the connector is sealed to the sensing cavity wall by laser welding. The data processing unit and the spectral detector of the optical path unit are connected by a shielded data cable, and the data cable is wrapped with an anti-electromagnetic interference sheath. The calibration and early warning unit and the data processing unit complete bidirectional data interaction through an industrial Ethernet interface. At the same time, the temperature and pressure sensor of the calibration and early warning unit is connected to the reserved interface inside the sealed sensing cavity through a high-temperature shielded signal cable.

[0037] The specific connection methods of the internal components of each unit are as follows: (a) Connection of internal components of the membrane degassing unit The microfiltration module at the oil inlet end is connected to the main oil inlet pipeline of the membrane degassing unit via a quick-release buckle. The differential pressure monitoring interface of the microfiltration module is connected to an external differential pressure sensor via a miniature copper tube. The differential pressure sensor is then connected to the signal acquisition module of the calibration and early warning unit via a signal line to achieve real-time feedback of the blockage status.

[0038] The oil inlet side of the PTFE hollow fiber membrane module is fixedly connected to the oil outlet of the microfiltration module by an integrated clamp. The gas outlet side of the membrane module is connected to the sealing flange through a PTFE hose. A conical sealing structure is used between the hose and the flange to ensure that there is no leakage during the gas transmission process.

[0039] (ii) Connection of internal components of the surface-enhanced Raman spectroscopy sensing unit The multi-component selective surface-enhanced Raman spectroscopy sensing substrate is fixed in the center of the sealed sensing cavity by a ceramic insulating bracket. The bottom of the bracket is fixed to the bottom of the sensing cavity by threading and sealing to ensure the stability of the substrate position and insulation from the cavity wall, thus avoiding electromagnetic interference.

[0040] The anti-reflective film on the inner wall of the sealed sensing cavity is combined with the cavity wall using a physical vapor deposition process. The observation window of the sensing cavity (for the connection between the fiber optic probe and the signal collection fiber) is airtightly connected to the cavity wall using a metal sealing gasket. The observation window is made of high-transmittance quartz glass to ensure the penetration of laser and Raman signals.

[0041] (III) Connection of internal components of the optical path unit The output end of the laser is connected to the input end of the fiber optic coupler via a fiber optic adapter. The adapter uses an FC / APC type low-loss connector to reduce energy loss during laser transmission. The output end of the fiber optic coupler is connected to the fiber optic probe via a single-mode transmission fiber. The probe end of the fiber optic probe is perpendicular to the surface of the sensing substrate and maintains a fixed distance from the substrate to ensure laser focusing effect.

[0042] The signal input end of the spectrometer is connected to the signal collection interface of the sealed sensing cavity through a multimode Raman signal collection fiber. A microlens is installed at the sensing cavity end of the collection fiber to improve the collection efficiency of the Raman signal. The signal output end of the spectrometer is connected to the signal receiving port of the data processing unit through an RS485 shielded data cable. Grounding terminals are installed at both ends of the data cable to eliminate electromagnetic interference in signal transmission.

[0043] (iv) Connection of internal components of the data processing unit The spectral preprocessing module, peak deconvolution module, and multi-model fusion quantitative module of the data processing unit are all integrated on the same industrial control motherboard. The modules interact internally through the onboard data bus, eliminating the need for additional external wiring and ensuring the timeliness of data processing.

[0044] The external signal interfaces of the data processing unit are divided into two categories: input and output. The input interface is connected to the spectral detector through a shielded data cable, and the output interface is connected to the calibration and early warning unit and the remote operation and maintenance platform through an industrial Ethernet, respectively, to realize bidirectional data transmission and sharing.

[0045] (v) Connection of internal components of the calibration and early warning unit The probes of the temperature and pressure sensors are embedded in the inner wall of the sealed sensing cavity through a sealed threaded interface. The signal terminals of the sensors are connected to the signal conditioning module of the calibration and early warning unit through a high-temperature shielded wire. The conditioning module amplifies and filters the original signal before transmitting it to the data compensation and calculation module.

[0046] The fault threshold database is integrated into the storage module of the calibration and early warning unit. The storage module is connected to the compensation calculation module through an internal data interface. The compensation calculation module compares the gas concentration data after temperature and pressure calibration with the database thresholds, and then connects to the audible and visual alarm or remote early warning platform through the early warning signal output interface to trigger the fault early warning.

[0047] Specifically, in this embodiment, the laser is a narrow-linewidth semiconductor laser with a linewidth not exceeding 0.001 pm, and the spectral detector is a charge-coupled device (CCD) with a pixel count of at least 2000×256. The narrow-linewidth semiconductor laser ensures the stability of the excitation wavelength, avoids spectral peak shifts caused by wavelength drift, and improves detection accuracy. The high-pixel CCCD can capture more subtle spectral features, identify weak characteristic peaks in low-concentration gases, and further enhance the accuracy of qualitative and quantitative gas analysis.

[0048] Specifically, in this embodiment, the spectral preprocessing module of the data processing unit includes a baseline correction submodule and a smoothing and denoising submodule. The smoothing and denoising submodule employs the Savitzky-Golay filtering algorithm. The baseline correction submodule of the spectral preprocessing module can eliminate the interference of baseline drift on the spectrum, while the denoising submodule, which uses a specific filtering algorithm, can completely preserve the characteristic peak shape while removing spectral noise, avoiding the loss of key information. This provides high-quality raw data for subsequent peak deconvolution and multi-model quantitative analysis, improving the reliability of data parsing.

[0049] Specifically, in this embodiment, the temperature sensor of the calibration and early warning unit has a measurement range of -25℃ to 70℃, and the pressure sensor has a measurement range of 0.05MPa to 0.35MPa. The fault threshold database is built according to the DL / T722-2018 standard. The temperature and pressure sensors adapted to the transformer's operating range can accurately collect environmental parameters to achieve environmental compensation for quantitative results, eliminating the impact of temperature and pressure fluctuations on the detection data. The fault threshold database built according to industry standards provides professional and authoritative judgment criteria for fault early warning, improving the accuracy of fault judgment and industry adaptability.

[0050] Specifically, in this embodiment, the inner wall of the sealed sensing cavity is coated with an anti-reflective film, the reflectivity of which does not exceed 5%, and the overall volume of the sensing cavity does not exceed 5 cm³. The anti-reflective film on the inner wall of the sealed sensing cavity can reduce the reflection and superposition of laser and Raman signals within the cavity, ensuring the purity of the acquired signal; the smaller cavity volume can shorten the time for gas to diffuse to the sensing substrate, improve the monitoring response speed, and at the same time reduce the installation space of the sensing unit, making it easier to arrange in the fault-prone and confined area inside the transformer tank.

[0051] Specifically, in this embodiment, a microfiltration component is provided at the oil inlet of the membrane degassing unit. The microfiltration component has a pore size of 1μm-2μm and is used to intercept solid impurities in the transformer oil. Adding a microfiltration component at the oil inlet of the membrane degassing unit can intercept solid impurities in the transformer oil, preventing impurities from clogging the membrane pores or damaging the membrane component. This ensures the long-term stable operation of the membrane degassing unit, maintains consistent oil-gas separation efficiency, extends the service life of the membrane component, and reduces the later maintenance costs of the monitoring system.

[0052] The microfiltration unit can be a customized Y-type cylindrical microfiltration unit, model WLY-TR-2.0, which is fully compatible with the DN15 oil inlet pipeline of the membrane degassing unit. Its pressure rating meets the working pressure requirements of the transformer tank, and its temperature range covers the transformer's operating temperature range. The unit housing is made of 304 stainless steel and adopts a Y-type flow channel design. Both the oil inlet and outlet are external threaded interfaces for easy connection with quick-release clips. The top of the housing integrates dual differential pressure monitoring interfaces, which can be connected to an external differential pressure transmitter. The interface specifications are as follows: M10×1 micro thread; a drain port with a ball valve at the bottom, the ball valve has a nominal diameter of 8mm, which allows for easy draining without stopping the machine, and the entire filter element can be replaced by removing the end cap; the filter element uses a PP pleated microfiltration membrane with a filtration pore size of 1.5μm, a filter element length of 100mm, an outer diameter of 30mm, and a 316L stainless steel porous support structure with high porosity and excellent mechanical strength, which can withstand long-term oil flow impact; the filter element sealing end uses an oil-resistant fluororubber O-ring to ensure no oil short-circuit leakage.

[0053] Specifically, in this embodiment, the model is based on selective adsorption and nonlinear enhancement effects, and is used in the multi-model fusion quantitative module of the data processing unit to accurately invert the concentration of characteristic peaks after deconvolution separation; the core equation of the model—the multi-gas Raman signal enhancement and separation quantitative model on the selective adsorption substrate—is expressed as follows: ; in: The value represents the net intensity (au) of the surface-enhanced Raman scattering (SERS) characteristic peak of the i-th characteristic gas at its characteristic Raman shift after preprocessing and deconvolution. The reference Raman enhancement factor (au·L / mg) of the i-th gas on the gold nanostar-bifunctional sensing substrate was obtained through a single-component calibration experiment. C i The concentration (mg / L) of the adsorbed phase of the i-th gas to be determined on the surface of the sensing substrate is related to the dissolved concentration in the oil via an adsorption isotherm. is the nonlinear enhancement attenuation coefficient (L / mg) for the i-th gas, used to characterize the effect that the enhancement efficiency decreases with increasing total adsorption amount due to changes in the occupancy of adsorption sites on the substrate surface or the coupling of the plasma resonance field; The dimensionless adsorption coefficient represents the competitive adsorption site of the j-th gas on the i-th gas, describing the intermolecular adsorption competition caused by the selectivity of the bifunctional group. The spectral cross-interference coefficient (au·L / mg) of the k-th gas on the characteristic peak of the i-th gas is used to correct the signal contribution caused by residual error of spectral deconvolution or physical interference. n represents the total number of types of characteristic gases to be measured; This is for random measurement noise and model residuals.

[0054] The derivation process of the process: 1. Starting point: The classical Raman scattering intensity I and the scattering molecule concentration C have a linear relationship within a certain range: I = G·C. In surface-enhanced Raman spectroscopy (SERS), G is the enhancement factor, which is usually much greater than 1.

[0055] 2. Introduction of Nonlinear Enhancement Effect: On the gold nanostar array substrate of this invention, the enhancement factor G is not constant. When multiple gas molecules selectively adsorb onto bifunctional sites, high concentrations or mixed adsorption may lead to: (a) saturation or weakening of the electromagnetic field enhancement efficiency in local "hot spot" regions due to increased molecular coverage; (b) slight alteration of local plasmon resonance conditions due to the adsorption of different molecules. Therefore, G is modified to be a concentration-dependent function: It adopts an exponential decay form. To simulate this nonlinear effect that decreases with increasing total effective adsorption, where Control the decay rate, Weighted contribution of different gases.

[0056] 3. Introducing a cross-interference term: Although a deconvolution algorithm is used, residual errors may exist in the complete separation of spectral peaks in complex gas mixtures. Furthermore, the physical adsorption of non-target molecules may also produce slight background interference. Therefore, a linear cross-interference term is added. This is to compensate for these systematic errors and improve quantitative robustness.

[0057] 4. Final Form: Combining steps 2 and 3, we obtain the SERS signal intensity model measured for a single gas in the gas mixture: .

[0058] Example: Assume the monitoring targets are H2 (i=1), CH4 (i=2), and C2H2 (i=3), a total of 3 gases (n=3).

[0059] 1. System Calibration (Preliminary Step): Measurements are performed in the laboratory using standard gases, both individually and in specific mixtures, at different concentrations. All parameters in the model are pre-determined using a nonlinear fitting algorithm. ,matrix and (Non-diagonal elements), and establish the adsorption phase concentration C. iWith the concentration of dissolved substances in transformer oil The transformation relationship (such as the modified form of Henry's Law).

[0060] 2. On-site monitoring and quantitative inversion: The system acquires spectra, and after preprocessing and peak deconvolution, obtains the measured intensities corresponding to the characteristic peaks of the three gases: , , (au).

[0061] The data processing unit calls the above model equations with calibrated parameters to form a system of nonlinear equations about C1, C2, and C3: ; The numerical solutions of the system of equations C1, C2, and C3 are obtained by using numerical iteration methods (such as the Newton-Raphson method).

[0062] According to C i and The calibration relationship is used to calculate the dissolved concentration in the oil.

[0063] The concentration results are sent to the calibration and early warning unit for temperature and pressure compensation and compared with the fault threshold.

[0064] Parameter description table:

[0065] Technical effects: 1. Improved quantitative accuracy: Compared to traditional linear superposition models (such as...) This model uses an exponential term. This more precisely describes the nonlinear physical reality of the SERS enhancement effect as a function of adsorption coverage, and explicitly demonstrates it through... and The competitive adsorption and spectral interference were quantified, thereby significantly improving the simultaneous quantitative accuracy of multi-component gases, especially under high concentration or complex mixing conditions.

[0066] 2. Enhanced robustness of the method: The model explicitly includes interference terms, making the algorithm more tolerant of incomplete spectral deconvolution and the presence of unknown weak interference, thus improving the robustness of analysis in complex field environments.

[0067] 3. Simplified and unified algorithm process: The model unifies the three core physicochemical processes of "nonlinear enhancement", "competitive adsorption" and "cross-interference" into a compact mathematical framework, replacing the complex process that may require multiple sub-models or empirical correction steps, making the core quantitative algorithm of the data processing unit simpler and more efficient.

[0068] 4. Revealing and utilizing physical mechanisms: model parameters ( These have clear physical or chemical meanings, and their calibration values ​​can serve as quantitative indicators of the performance of the sensing substrate (enhancement capability, selectivity, and anti-saturation), guiding substrate optimization and realizing an upgrade from phenomenon description to mechanism modeling.

[0069] Working principle and process: 1. Signal acquisition: The optical path unit acquires the raw SERS spectrum and transmits it to the data processing unit.

[0070] 2. Spectral preprocessing: The spectral preprocessing module performs baseline correction and Savitzky-Golay filtering to obtain a clean spectrum.

[0071] 3. Feature Peak Deconvolution: The peak deconvolution module separates overlapping peaks and outputs the net intensity of the feature peaks corresponding to each feature gas. .

[0072] 4. Model Quantitative Calculation: The multi-model fusion quantitative module calls the core quantitative model proposed in this invention. As input, combined with the pre-stored calibration parameter set within the module ( By solving the nonlinear equations, the adsorption phase concentration C of each gas can be derived. i .

[0073] 5. Concentration Conversion and Compensation: Based on the pre-stored adsorption-dissolution equilibrium relationship, C... i This is converted to the dissolved concentration in oil. The calibration and early warning unit simultaneously reads the temperature and pressure sensor data, performs environmental compensation on this concentration, and obtains the final gas concentration under standard conditions.

[0074] 6. Fault diagnosis: The compensated concentration is compared with the fault threshold database to achieve fault early warning.

[0075] Working principle: This scheme is an integrated monitoring system based on surface-enhanced Raman spectroscopy, combined with in-situ oil and gas separation, multi-algorithm data analysis, and environmental compensation. Its core working principle can be divided into four key aspects: 1. Selective adsorption and Raman signal enhancement Gold nanostar arrays based on silicon arrays generate localized surface plasmon resonance effects through their unique tip structure, which can significantly amplify the Raman scattering signals of characteristic gas molecules. The thiol and amino bifunctional groups grafted on the substrate surface can form specific interactions with target gas molecules, achieving selective adsorption of characteristic fault gases in transformer oil and filtering out interference from non-target components, laying the foundation for subsequent accurate detection.

[0076] 2. In-situ oil-gas separation The membrane degassing unit uses a polytetrafluoroethylene hollow fiber membrane module, which utilizes its oil-repellent and gas-permeable properties to directly separate the oil phase and gas phase inside the transformer tank. The characteristic gases dissolved in the oil can penetrate the membrane module and enter the sealed sensing cavity, while the insulating oil and solid impurities in the oil are effectively intercepted. The microfiltration module at the oil inlet can further block solid particles, avoid membrane pore blockage, ensure the continuous stability of the separation process, and achieve in-situ gas acquisition without sampling.

[0077] 3. Spectral Acquisition and Data Analysis The excitation laser emitted by the narrow-linewidth near-infrared laser in the optical path unit is transmitted to the sensing substrate through optical fiber and excites the adsorbed gas to generate Raman scattering signals. The high-pixel spectral detector can accurately capture these signals. The acquired raw spectrum is first preprocessed to eliminate interference such as baseline correction and smoothing and noise reduction. Then, the overlapping characteristic peaks are separated by the deconvolution algorithm. Finally, the qualitative identification and quantitative analysis of multi-component gases are completed by relying on the partial least squares discriminant analysis and support vector machine fusion algorithm, which solves the problems of overlapping multi-gas peaks and difficulty in interpreting low-concentration signals.

[0078] 4. Environmental compensation and fault early warning The temperature and pressure sensors of the calibration and early warning unit collect the temperature and pressure parameters of the transformer's operating environment in real time, perform environmental compensation on the quantitative analysis results, and eliminate the detection deviation caused by temperature and pressure fluctuations. At the same time, combined with the fault threshold database built according to industry standards, the compensated gas concentration data is compared with the threshold. Once it exceeds the safe range, the fault early warning is automatically triggered, realizing closed-loop management from data detection to fault diagnosis.

[0079] How to use: 1. Preliminary preparations and system installation First, the multi-component selective surface-enhanced Raman spectroscopy sensing substrate was prepared. Following the established process, a gold nanostar array was grown on a silicon-based array and grafted with bifunctional groups. Then, a complete monitoring system was built. The sealed sensing cavity of the integrated sensing substrate was sealed and connected to the membrane degassing unit and the optical path unit. The data processing unit and the calibration and early warning unit were connected in circuitry and signal configuration. Finally, the monitoring probe was pre-embedded in fault-prone areas within the transformer tank, such as winding joints and the core, ensuring communication between the oil inlet of the membrane degassing unit and the inside of the tank. Simultaneously, the temperature and pressure sensors were arranged, and the parameters of the fault threshold database were adjusted.

[0080] 2. System startup and real-time monitoring After the monitoring system is started, the transformer oil first passes through the microfiltration component to intercept solid impurities, and then enters the membrane degassing unit to complete in-situ oil-gas separation. The characteristic gas permeates to the surface of the sensing substrate and is selectively adsorbed. The optical path unit starts the laser to emit excitation laser, and the spectral detector simultaneously collects the Raman scattering spectrum and transmits the raw spectral data to the data processing unit.

[0081] 3. Data Processing and Result Output The data processing unit performs baseline correction, smoothing and denoising, and peak deconvolution processing on the original spectrum in sequence, and then completes the qualitative and quantitative analysis of the gas through a multi-model fusion algorithm. The calibration and early warning unit compensates for the quantitative results by combining real-time temperature and pressure data, and finally outputs accurate gas composition and concentration data. If the data exceeds the fault threshold, an early warning is automatically triggered to remind maintenance personnel to troubleshoot the fault in time.

[0082] 4. Post-maintenance Regularly check the cleanliness of the microfiltration components and replace clogged filters in a timely manner; perform performance testing on the membrane components of the membrane degassing unit to ensure oil-gas separation efficiency; and simultaneously verify the laser wavelength stability and spectral detector sensitivity of the optical path unit to ensure long-term stable operation of the system.

[0083] In summary, the in-situ monitoring method and system for dissolved gases in transformer oil using surface-enhanced Raman spectroscopy provided in this embodiment has the following advantages: 1. Solving the problem of lag in traditional monitoring: By pre-embedded probes combined with the in-situ oil-gas separation design of the membrane degassing unit, online in-situ monitoring of dissolved gases in transformer oil is realized. There is no need for offline sampling and testing. Gas composition data can be obtained in real time, which saves time for the early detection of internal transformer faults and avoids the expansion of faults due to monitoring lag.

[0084] 2. Enhanced monitoring sensitivity and specificity: The structural design of the gold nanostar array enhances the Raman signal, enabling even low-concentration fault characteristic gases to generate identifiable Raman signals; the grafting of bifunctional groups endows the sensing substrate with selective adsorption capabilities for target gases, effectively filtering out interference from non-target components and impurities in the oil, thus significantly improving the monitoring sensitivity and specificity.

[0085] 3. Ensuring the accuracy of detection data: The combination of a narrow-linewidth laser and a high-pixel spectral detector ensures the stability and accuracy of Raman signal acquisition; multi-step spectral preprocessing eliminates interference such as baseline drift and signal noise; peak deconvolution technology solves the problem of overlapping characteristic peaks; and multi-algorithm fusion analysis improves the reliability of qualitative and quantitative analysis of multi-component gases; at the same time, the temperature and pressure environment compensation mechanism eliminates the impact of transformer operating environment fluctuations on detection results, further ensuring the accuracy of the data.

[0086] 4. Enhanced system practicality and durability: The microfiltration component at the oil inlet can intercept solid impurities, preventing membrane module clogging or damage, extending the service life of the membrane degassing unit, and reducing the frequency of later system maintenance; the anti-reflective membrane design of the sealed sensing cavity ensures the purity of signal acquisition, while the miniaturized cavity structure facilitates placement in the narrow, fault-prone areas of the transformer tank, improving the system's field adaptability; the fault threshold database built according to industry standards provides professional and authoritative judgment criteria for fault early warning, enhancing the engineering practicality of the technical solution.

[0087] 5. Forming an integrated monitoring closed loop: From in-situ oil and gas separation and signal enhancement acquisition to accurate data analysis, environmental compensation, and fault early warning, the entire technology forms a complete monitoring closed loop, which can realize the fully automated processing from gas detection to fault diagnosis. It provides integrated technical support for transformer operation and maintenance, reduces the workload of operation and maintenance personnel, and improves the operation and maintenance management level of power grid equipment.

[0088] In addition, this embodiment also provides a specific example of in-situ monitoring of dissolved gas in 220kV power transformer oil using surface-enhanced Raman spectroscopy.

[0089] I. Technical Solution This example addresses the need for dissolved gas monitoring in the oil of a 220kV main transformer winding hotspot area. It establishes a suitable in-situ monitoring system and implements a monitoring method, as detailed below: (I) Monitoring System Setup Membrane degassing unit: The oil inlet is connected to the hot spot area of ​​the transformer oil tank winding, and a micro-filter is installed at the oil inlet to intercept solid impurities in the oil; the core uses a polytetrafluoroethylene hollow fiber membrane module to achieve in-situ separation of dissolved gases in the oil, ensuring that the gases can smoothly enter the subsequent detection process.

[0090] Surface-enhanced Raman spectroscopy sensing unit: It is equipped with a miniaturized sealed sensing cavity, the inner wall of which is coated with an anti-reflection film to reduce signal reflection interference; a multi-component selective sensing substrate is fixed in the cavity. The substrate uses a specific periodic silicon-based array as a carrier, and a gold nanostar array is grown on the surface. Thiol groups and amino functional groups are grafted sequentially to form characteristic gas-specific adsorption sites.

[0091] Optical path unit: A narrow linewidth semiconductor laser is selected as the excitation source, which is connected to the fiber optic probe via a fiber optic coupler and faces the sensing substrate; a high-pixel charge-coupled detector is used as the spectral detector, and the Raman signal collection fiber is connected to the sealed sensing cavity to realize precise laser excitation and efficient Raman signal acquisition.

[0092] Data processing unit: Built-in spectral preprocessing module (including baseline correction submodule and Savitzky-Golay filtering smoothing and denoising submodule), peak deconvolution module and multi-model fusion quantitative module, which can complete the optimization processing of raw spectra and qualitative and quantitative analysis of multi-component gases.

[0093] Calibration and early warning unit: Equipped with a wide-range temperature and pressure sensor to collect environmental parameters, and at the same time, a fault threshold database is built according to the DL / T722-2018 standard to realize environmental compensation of detection data and intelligent fault early warning.

[0094] (II) Implementation of Monitoring Methods Substrate preparation: First, a gold nanostar array is directionally grown on the surface of a silicon-based array to construct a strong Raman-enhanced "hot spot". Then, thiol and amino functional groups are grafted sequentially to complete the preparation of the characteristic gas selective adsorption structure.

[0095] In-situ oil-gas separation and adsorption: The monitoring probe with integrated sensing substrate is pre-embedded in the hot spot area of ​​the transformer winding. The transformer oil enters the membrane degassing unit after being filtered by the microfiltration component. The characteristic gases dissolved in the oil penetrate the membrane component and enter the sealed sensing cavity, where they are selectively adsorbed by the functional groups on the surface of the sensing substrate.

[0096] Raman spectroscopy acquisition: The optical path unit emits a near-infrared excitation laser, controls the laser power to balance the signal intensity and substrate protection, acquires Raman scattering spectra covering the peak position range of characteristic gases in transformer oil, and transmits them to the data processing unit.

[0097] Spectral analysis and quantification: First, baseline correction and denoising preprocessing are performed on the original spectrum. Then, overlapping feature peaks are separated by deconvolution algorithm. Finally, partial least squares discriminant analysis combined with support vector machine algorithm is used to complete the qualitative identification and quantitative calculation of multi-feature gases.

[0098] Environmental compensation and fault early warning: The quantitative results are compensated for by temperature and pressure sensor data to eliminate the influence of temperature and pressure fluctuations. The compensated data is then compared with the fault threshold database to output the gas component concentration and trigger the corresponding early warning.

[0099] II. Working Principle Signal enhancement and selective adsorption principle: The tip structure of the gold nanostar array can excite local surface plasmon resonance, which greatly enhances the Raman scattering signal of the adsorbed gas; the grafted bifunctional groups can specifically bind with characteristic gases such as H2 and C2H2, thereby achieving selective capture of the target gas and eliminating interference from non-target components.

[0100] In-situ oil-gas separation principle: The polytetrafluoroethylene hollow fiber membrane has oil-repellent and gas-permeable properties, allowing only dissolved gases in the oil to pass through while blocking the insulating oil. The microfiltration component simultaneously intercepts solid impurities, avoiding membrane pore blockage, thereby completing in-situ oil-gas separation without sampling inside the transformer.

[0101] Spectral analysis principle: Narrow linewidth lasers can reduce fluorescence background interference, and high-pixel detectors can capture weak Raman signals; the spectral preprocessing module eliminates baseline drift and noise, the peak deconvolution module splits overlapping feature peaks, and multi-algorithm fusion analysis combines the advantages of two algorithms to improve the accuracy of multi-component gas analysis.

[0102] Environmental compensation and early warning principle: Changes in transformer temperature and pressure affect gas state and detection signal. After collecting parameters, temperature and pressure sensors compensate for quantitative results and correct environmental deviations. The fault threshold database sets a safe range according to industry standards. When the concentration exceeds the standard, an early warning is automatically triggered to achieve early fault identification.

[0103] III. Experimental Data To verify the effectiveness of the scheme, four groups of simulated oil samples (normal oil sample, low-energy discharge fault oil sample, medium-temperature overheating fault oil sample, and high-temperature overheating fault oil sample) were selected for testing, and a 30-day continuous stability test was conducted. The specific data are as follows: Qualitative identification data: The characteristic gas components of all four oil samples were accurately identified without any misidentification of components. After deconvolution processing, the overlapping C2H2 and C2H4 characteristic peaks could be completely distinguished, and the identification results were consistent with those of the standard gas chromatography method.

[0104] Quantitative comparison data: The quantitative results of each characteristic gas have a small deviation from the detection values ​​of gas chromatography. After temperature and pressure compensation, the deviation is further reduced, meeting the accuracy requirements of on-site monitoring. Even low concentrations of H2 can be effectively quantified.

[0105] Response timeliness data: The overall response time from oil and gas separation to the output of the final detection result is relatively short. Compared with traditional offline detection, the timeliness is significantly improved, which can meet the needs of real-time monitoring.

[0106] Stability data: During 30 days of continuous operation, the separation efficiency of the membrane degassing unit showed no significant decline, the Raman enhancement effect of the sensing substrate remained stable, the data processing unit analysis results showed minimal fluctuations, and the system had no records of fault shutdowns.

[0107] IV. Technical Effects The design of the pre-embedded probe combined with the membrane degassing unit enables in-situ real-time monitoring of dissolved gases in oil, completely solving the lag problem of traditional offline sampling and providing a time advantage for early warning of transformer faults.

[0108] The structural design of the gold nanostar array enhances the Raman signal enhancement effect and improves the detection capability of low-concentration fault gases. The grafting of bifunctional groups endows the substrate with selective adsorption capability, reduces the interference of impurities and non-target gases in the oil, and ensures the sensitivity and specificity of monitoring.

[0109] Multi-step spectral preprocessing eliminates signal interference, peak deconvolution technology solves the problem of overlapping characteristic peaks, multi-algorithm fusion analysis improves quantitative and qualitative accuracy, and combined with temperature and pressure compensation mechanism, further ensures the accuracy and reliability of detection data.

[0110] The microfiltration module extends the service life of the membrane degassing unit and reduces system maintenance costs; the anti-reflective membrane in the sealed sensing cavity ensures the purity of signal acquisition, and the miniaturized design is suitable for transformer installation in confined spaces; the industry standard database ensures the authority of the early warning results and improves the engineering practicality of the solution.

[0111] The above are merely preferred embodiments of the present invention and are not intended to limit the implementation methods and protection scope of the present invention. Those skilled in the art should recognize that any equivalent substitutions and obvious changes made based on the description and illustrations of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for in-situ monitoring of dissolved gases in transformer oil using surface-enhanced Raman spectroscopy, characterized in that, Includes the following steps: Step 1: Prepare a multi-component selective surface-enhanced Raman spectroscopy sensing substrate, grow a gold nanostar array on the silicon-based array surface, and sequentially graft thiol functional groups and amino functional groups to form selective adsorption sites for characteristic gases in transformer oil. Step 2: The monitoring probe integrating the surface-enhanced Raman spectroscopy sensing substrate is pre-embedded in the fault-prone area inside the transformer oil tank. The dissolved gas in the transformer oil is separated in situ through the membrane degassing unit, so that the characteristic gas permeates to the surface of the surface-enhanced Raman spectroscopy sensing substrate. Step 3: Emit an excitation laser to the surface-enhanced Raman spectroscopy sensing substrate through the optical path unit to collect the surface-enhanced Raman scattering spectrum of the characteristic gas; Step 4: Preprocess the acquired raw spectra, separate overlapping feature peaks using a deconvolution algorithm, and then use partial least squares discriminant analysis combined with support vector machine algorithm to achieve qualitative identification and quantitative analysis of multi-feature gases; Step 5: Perform environmental compensation on the quantitative results through the temperature-pressure calibration module, output gas component and concentration data, and complete fault warning.

2. The in-situ monitoring method for dissolved gases in transformer oil using surface-enhanced Raman spectroscopy according to claim 1, characterized in that, The gold nanostar array in step 1 has a particle size of 120nm-180nm and a tip curvature radius of 8nm-12nm, and the silicon-based array has a period of 1.8μm-2.2μm.

3. The in-situ monitoring method for dissolved gases in transformer oil using surface-enhanced Raman spectroscopy according to claim 1, characterized in that, The membrane module of the membrane degassing unit in step 2 is a polytetrafluoroethylene hollow fiber membrane with a pore size of 0.15μm-0.25μm and an oil-gas separation efficiency of not less than 85%.

4. The in-situ monitoring method for dissolved gases in transformer oil using surface-enhanced Raman spectroscopy according to claim 1, characterized in that, The excitation laser used in step 3 is a 785nm near-infrared laser with an output power of 40mW-60mW, and the detection range of the surface-enhanced Raman scattering spectrum is 200cm. -1 -4000cm -1 .

5. A surface-enhanced Raman spectroscopy in-situ monitoring system for dissolved gases in transformer oil, characterized in that, The method for in-situ monitoring of dissolved gases in transformer oil using surface-enhanced Raman spectroscopy as described in any one of claims 1-4 includes a membrane degassing unit, a surface-enhanced Raman spectroscopy sensing unit, an optical path unit, a data processing unit, and a calibration and early warning unit. The oil inlet of the membrane degassing unit is connected to the inside of the transformer oil tank, and the gas outlet is sealed and connected to the surface-enhanced Raman spectroscopy sensing unit. The surface-enhanced Raman spectroscopy sensing unit includes the multi-component selective surface-enhanced Raman spectroscopy sensing substrate and the sealed sensing cavity as described in claim 1, wherein the surface-enhanced Raman spectroscopy sensing substrate is fixed inside the sealed sensing cavity; The optical path unit includes a laser, an optical fiber coupler, an optical fiber probe, and a spectral detector. The output end of the laser is connected to the optical fiber probe through the optical fiber coupler. The optical fiber probe faces the surface-enhanced Raman spectroscopy sensing substrate. The spectral detector is connected to the sealed sensing cavity through a Raman signal collecting optical fiber. The data processing unit has a built-in spectral preprocessing module, a peak deconvolution module, and a multi-model fusion quantitative module. The calibration and early warning unit includes a temperature sensor, a pressure sensor, and a fault threshold database.

6. The in-situ monitoring system for dissolved gases in transformer oil using surface-enhanced Raman spectroscopy according to claim 5, characterized in that, The laser is a narrow linewidth semiconductor laser with a linewidth not exceeding 0.001 pm, and the spectral detector is a charge-coupled detector with a pixel count of not less than 2000×256.

7. The in-situ monitoring system for dissolved gases in transformer oil using surface-enhanced Raman spectroscopy according to claim 5, characterized in that, The spectral preprocessing module of the data processing unit includes a baseline correction submodule and a smoothing and denoising submodule. The smoothing and denoising submodule uses the Savitzky-Golay filtering algorithm.

8. The in-situ monitoring system for dissolved gases in transformer oil using surface-enhanced Raman spectroscopy according to claim 5, characterized in that, The temperature sensor of the calibration and early warning unit has a measurement range of -25℃ to 70℃, and the pressure sensor has a measurement range of 0.05MPa to 0.35MPa. The fault threshold database is built according to the DL / T722-2018 standard.

9. The in-situ monitoring system for dissolved gases in transformer oil using surface-enhanced Raman spectroscopy according to claim 5, characterized in that, The inner wall of the sealed sensing cavity is coated with an anti-reflective film, the reflectivity of which does not exceed 5%, and the overall volume of the sensing cavity does not exceed 5 cm³.

10. The in-situ monitoring system for dissolved gases in transformer oil using surface-enhanced Raman spectroscopy according to claim 5, characterized in that, The oil inlet of the membrane degassing unit is equipped with a microfiltration component with a pore size of 1μm-2μm, which is used to intercept solid impurities in the transformer oil.

Images