A transformer oil detection method based on polarized light and ultraviolet light

Abstract

The application discloses a transformer oil detection method based on polarized light and ultraviolet light, and relates to the technical field of electrical equipment insulating oil detection.The method comprises the following steps: using a polarized light source to irradiate a stable flowing oil sample to be detected, and acquiring polarized light field distribution data after transmission and scattering; using an ultraviolet light source to irradiate the oil sample, and acquiring characteristic fluorescence spectrum and spatial intensity distribution data generated by excitation; respectively establishing mapping models of the polarized light field distribution data and water content and suspended particulate matter in the oil, and mapping models of characteristic fluorescence data and aging products and dissolved fault gases, so as to quantitatively analyze various impurity parameters; and finally, fusing the parameters to construct a comprehensive evaluation index of transformer oil state, and generating a diagnosis report of insulating performance and aging degree according to a threshold standard.The application realizes high-precision, spatially-resolved synchronous detection and comprehensive evaluation of various forms of pollutants in transformer oil.

Description

Technical Field

[0001] This invention belongs to the field of electrical equipment insulating oil testing technology, specifically a transformer oil testing method based on polarized light and ultraviolet light. Background Technology

[0002] The insulation condition assessment of transformer oil relies on the accurate detection of its internal suspended and dissolved impurities. Existing laser detection technologies for suspended particles mainly estimate particle counts through light intensity attenuation or scattering signals, but they struggle to effectively distinguish between particle morphology and interference from tiny water droplets, resulting in insufficient resolution of complex contaminants. For trace moisture, the commonly used coulometric method, while highly accurate, is an offline analysis and cannot achieve simultaneous, in-situ measurement with particulate matter. Existing methods are mostly limited to discrete measurements of single physical quantities, lacking comprehensive optical information that can simultaneously reflect particle characteristics and moisture content.

[0003] In the analysis of soluble components, gas chromatography has a long analysis cycle and is not suitable for rapid on-site monitoring. Traditional fluorescence spectroscopy identifies substances based on emission spectra, but only records the change of spectral intensity with wavelength, lacking information on the spatial distribution of fluorescence signals within the sample. This makes it impossible to determine whether characteristic fluorescent substances are locally enriched or unevenly distributed, limiting the possibility of assessing the location of faults or the heterogeneity of aging processes from a spatial perspective.

[0004] There is a need to develop a detection technology capable of simultaneously and in-situ acquiring multidimensional information on various forms of impurities in oil. This technology needs to be able to extract richer features from optical responses to simultaneously achieve quantitative analysis and spatial distribution analysis of suspended particulate matter, trace moisture, dissolved gases, and aging products, overcoming the shortcomings of existing methods in terms of single detection dimensions and insufficient information fusion. Summary of the Invention

[0005] This invention aims to solve at least one of the technical problems existing in the prior art; Therefore, this invention proposes a transformer oil detection method based on polarized light and ultraviolet light, comprising: A polarized light source is used to irradiate a transformer oil sample under stable flow to obtain polarized light field distribution data formed by the transmission and scattering of the transformer oil sample. The transformer oil sample to be tested is irradiated with an ultraviolet light source to excite specific components in the transformer oil sample to produce characteristic fluorescence, and the fluorescence spectrum and spatial intensity distribution data are obtained. A mapping correlation model is established between polarized light field distribution data and suspended particulate matter and water content in oil. Based on the mapping correlation model, the polarized light field distribution data is analyzed to quantitatively resolve the parameters of suspended particulate matter concentration and trace water content in oil. A mapping correlation model is established between characteristic fluorescence spectra and characteristic gases of solubility failures and aging products in oil. Based on the mapping correlation model, the fluorescence spectra and spatial intensity distribution data are analyzed to identify and quantify the types and concentrations of characteristic gases of solubility failures and the types and contents of aging products in oil. The parameters of suspended particulate matter concentration and trace water content in the oil are fused with the types and concentrations of dissolved fault characteristic gases and the types and contents of aging products to construct a comprehensive evaluation index for transformer oil condition. Based on the comprehensive evaluation index of transformer oil condition and combined with the preset transformer oil condition threshold standard, a diagnostic report on the insulation performance and aging degree of the transformer oil sample to be tested is generated.

[0006] Furthermore, the step of irradiating the transformer oil sample under test with a polarized light source in a stable flowing state, and obtaining the polarized light field distribution data formed by the transmission and scattering of the transformer oil sample under test, specifically includes: The transformer oil sample to be tested is passed through an optically transparent detection flow cell at a preset constant flow rate. A linearly polarized light source of a single wavelength or narrow band is used to irradiate the transformer oil sample to be tested in the detection flow cell at a preset incident angle. Multiple polarization state sensors are arranged at multiple preset spatial angle positions in the detection flow cell to collect polarized light signals in the transmission direction and scattered polarized light signals at at least two specific scattering angles, respectively. Each of the polarization state sensors synchronously records the Stokes parameters of the received optical signal, which include light intensity, degree of polarization, polarization azimuth angle, and ellipticity angle. Stokes parameters collected from different spatial angles are aligned and integrated according to timestamps and spatial locations to form polarization field distribution data that describes the spatiotemporal evolution of the polarization field of the transformer oil sample under test during the detection period.

[0007] Furthermore, the step of irradiating the transformer oil sample with an ultraviolet light source to excite specific components in the sample to produce characteristic fluorescence and obtain fluorescence spectrum and spatial intensity distribution data specifically includes: A pulsed or continuous ultraviolet light source is used, wherein the center wavelength of the ultraviolet light source is located in the band that can excite the aromatic hydrocarbon compounds and fault characteristic gas derivatives in the transformer oil to produce fluorescence; The ultraviolet light source beam is focused onto the detection area in the detection flow cell that overlaps with or is adjacent to the polarized light irradiation area. The excited fluorescence signal is collected by using a spectrometer in the detection area at a certain angle to the incident light, and the resolution of the spectrometer is sufficient to distinguish the fluorescence characteristic peaks of different components. Simultaneously, an area array detector is used to acquire images of the spatial distribution of fluorescence generated in the detection area under ultraviolet light irradiation; The time-series fluorescence spectral data acquired by the spectrometer and the time-series fluorescence spatial distribution image acquired by the area array detector are synchronized in time and registered in space to obtain fluorescence spectral and spatial intensity distribution data containing spectral and two-dimensional spatial information.

[0008] Furthermore, the establishment of a mapping correlation model between polarized light field distribution data and the content of suspended particulate matter and water in oil, and the analysis of the polarized light field distribution data based on the mapping correlation model to quantitatively resolve the parameters of suspended particulate matter concentration and trace water content in oil, specifically includes: A series of standard transformer oil samples with known gradient concentrations of suspended particulate matter and known trace water content were prepared. Using the same detection conditions as those used to acquire the polarization field distribution data of the transformer oil sample to be tested, standard polarization field distribution data of each of the standard transformer oil samples was acquired. The variation law of Stokes parameter in the standard polarized light field distribution data was analyzed, and a set of characteristic parameters sensitive to the changes in suspended particulate matter concentration and moisture content was extracted. The set of characteristic parameters includes the attenuation rate of polarization degree of transmitted light, the polarization azimuth angle offset of scattered light at a specific scattering angle, and the spatial distribution gradient of the light field depolarization index. Based on the set of feature parameters and the known concentration of suspended particulate matter and trace water content, a mapping and correlation model from the set of feature parameters to the concentration of suspended particulate matter and trace water content is established by training through multivariate nonlinear regression or machine learning algorithms. The polarization field distribution data of the transformer oil sample to be tested is input into the mapping correlation model to calculate the quantitative analysis results of the suspended particulate matter concentration and trace water content parameters in the transformer oil sample to be tested.

[0009] Furthermore, the establishment of a mapping correlation model between characteristic fluorescence spectra and characteristic gases of dissolved failures and aging products in oil, and the analysis of fluorescence spectra and spatial intensity distribution data based on the mapping correlation model to identify and quantify the types and concentrations of characteristic gases of dissolved failures and the types and contents of aging products in oil, specifically includes: A series of standard transformer oil samples containing single or mixed known types and concentrations of soluble fault characteristic gases and aging products were prepared. Using the same detection conditions as those used to obtain the fluorescence spectrum and spatial intensity distribution data of the transformer oil sample to be tested, standard fluorescence spectrum and spatial intensity distribution data of each of the standard transformer oil samples were obtained. Feature extraction is performed on the standard fluorescence spectrum and spatial intensity distribution data to obtain the characteristic fluorescence spectral fingerprint and characteristic fluorescence spatial distribution pattern of each known component. The characteristic fluorescence spectral fingerprint includes the characteristic peak position, full width at half maximum (FWHM), and peak area ratio. The characteristic fluorescence spatial distribution pattern includes the fluorescence intensity spatial uniformity index and diffusion profile. Using spectral unmixing and pattern recognition algorithms, a mapping and correlation model is established to extract the contribution of characteristic fluorescence spectral fingerprints and the similarity of characteristic fluorescence spatial distribution patterns of each component from the fluorescence spectra and spatial intensity distribution data to be tested. The contribution and similarity are proportional to the component concentration and content. The fluorescence spectrum and spatial intensity distribution data of the transformer oil sample to be tested are input into the mapping correlation model, and the identification and quantification results of the type and concentration of dissolved fault characteristic gases and the types and contents of aging products in the oil are obtained by solving the model.

[0010] Furthermore, the process of fusing the concentration of suspended particulate matter and trace water content in the oil with the types and concentrations of dissolved fault characteristic gases and the types and contents of aging products to construct a comprehensive evaluation index for transformer oil condition specifically includes: The quantitative analysis results of the suspended particulate matter concentration, the quantitative analysis results of the trace water content parameters, and the identification and quantification results of the type and concentration of the characteristic gas of the dissolved failure and the type and content of the aging products are uniformly converted into a preset standardized dimension space. Within the standardized dimension space, each parameter is assigned a weighting coefficient based on its importance in the assessment of transformer oil insulation performance and aging status. Based on the standardized value of each parameter and its corresponding weight coefficient, a weighted fusion algorithm is used to calculate a comprehensive preliminary score of transformer oil condition. The optical field spatial uniformity index in the polarization optical field distribution data and the fluorescence diffusion profile index in the fluorescence spatial distribution data are introduced as correction factors for the preliminary score of transformer oil condition, and the preliminary score is dynamically corrected. The dynamically corrected score will be used as the comprehensive evaluation index for transformer oil condition.

[0011] Furthermore, the step of generating a diagnostic report on the insulation performance and aging degree of the transformer oil sample based on the comprehensive evaluation index of transformer oil condition and a preset threshold standard for transformer oil condition specifically includes: Multiple transformer oil condition levels are preset, each level corresponds to a threshold range of a comprehensive evaluation index for transformer oil condition, and is associated with different insulation performance descriptions and aging degree descriptions; The transformer oil condition comprehensive evaluation index of the transformer oil sample to be tested is compared with the threshold range of the transformer oil condition comprehensive evaluation index to determine the condition level of the transformer oil sample to be tested. Extract the preset insulation performance description and aging degree description belonging to the aforementioned state level as the basis for the diagnostic report; The analysis results of the polarization field distribution data are used to extract the type tendency information of the main pollutants, and the analysis results of the fluorescence spectrum and spatial intensity distribution data are used to extract the type tendency information of the main failure gases or aging products. The type and tendency information of the main pollutants and the type and tendency information of the main fault gases or aging products, combined with the specific values ​​in the quantitative analysis results and identification and quantification results, are filled into the basic content of the diagnostic report to generate a detailed diagnostic report that includes the state level, key parameter values ​​and material tendency analysis.

[0012] Furthermore, the Stokes parameters collected from different spatial angles are aligned and integrated according to timestamps and spatial locations to form polarization field distribution data describing the spatiotemporal evolution of the polarization field of the transformer oil sample under test during the detection period. Specifically, this includes: Add high-precision timestamps and sensor spatial location codes to the Stokes parameter data streams acquired by each polarization state sensor; Establish a data cube structure with time as the horizontal axis and spatial location and Stokes parameters as the dimensions; Synchronize the Stokes parameter sequences from all polarization state sensors to a unified time grid according to the timestamps; Based on the sensor spatial location encoding, the synchronized Stokes parameters are filled into the corresponding positions of the data cube structure; Spatial interpolation is performed on the data cube structure to obtain continuously distributed polarized light field information at a preset spatial resolution, forming complete polarized light field distribution data.

[0013] Furthermore, the time-series fluorescence spectral data acquired by the spectrometer and the time-series fluorescence spatial distribution image acquired by the area array detector are time-synchronized and spatially registered to obtain fluorescence spectral and spatial intensity distribution data containing spectral and two-dimensional spatial information, specifically including: Add a timestamp based on the same time reference to each spectral data frame acquired by the spectrometer and each image data frame acquired by the area array detector. The physical position of the optical fiber probe of the spectrometer in the image of the array detector is used as the reference point for spatial registration. Based on timestamp alignment, the spectral data and image data corresponding to each time point are associated to form a "spectral-image" data pair with time sequence. Geometric correction and intensity calibration are performed on the fluorescence spatial distribution image acquired by the area array detector to ensure that the spatial coordinates and fluorescence intensity values ​​of each pixel in the image are accurate. The spatial coordinates and intensity information of each pixel in the calibrated image are associated and extended with the spectral data of its corresponding time point to construct a dataset containing time, two-dimensional spatial coordinates, and fluorescence spectral intensity distribution at the corresponding spatial location point, namely fluorescence spectrum and spatial intensity distribution data.

[0014] Furthermore, the spatial uniformity index of the polarization field distribution data and the fluorescence diffusion profile index of the fluorescence spatial distribution data are introduced as correction factors for the preliminary transformer oil condition score, and the preliminary score is dynamically corrected, specifically including: The variance of transmitted light intensity or degree of polarization on the detection section is calculated from the polarized light field distribution data and used as an index of the spatial uniformity of the light field to characterize the uniformity of the oil. The fluorescence intensity decay curve characteristic parameter from the excitation center outward is extracted from the fluorescence spatial distribution data and used as a fluorescence diffusion profile index to characterize the diffusion or distribution of dissolved substances. An empirical relationship is established between the light field spatial uniformity index and the fluorescence diffusion profile index and the stability of the oil condition. When the light field spatial uniformity index is below the threshold and the fluorescence diffusion profile index shows rapid decay, the oil condition is determined to be relatively stable. Based on the aforementioned empirical relationship, the optical field spatial uniformity index and the fluorescence diffusion profile index are converted into correction coefficients for the preliminary score of the transformer oil condition. The initial transformer oil condition score is corrected by multiplying the correction coefficient to obtain the final comprehensive evaluation index of transformer oil condition calibrated by stability factors.

[0015] Compared with the prior art, the beneficial effects of the present invention are: This method employs a polarized light source to illuminate and analyze the polarized light field distribution data after transmission and scattering, rather than simply measuring light intensity. The spatial distribution of polarization states exhibits a specific response to the anisotropy of particulate matter, surface roughness, and the birefringence effect of water droplets. This approach can separate unique polarization modulation information caused by particle shape, size distribution, and trace moisture from complex optical signals, thereby achieving simultaneous and highly selective quantitative analysis of suspended particulate matter concentration, particle size distribution characteristics, and trace moisture content, improving the identification capability in the presence of mixed pollutants.

[0016] This method uses an ultraviolet light source to excite and simultaneously acquire fluorescence spectra and their intensity distribution data within the sample space. It not only captures the characteristic wavelengths emitted by fault-specific gases or aging product molecules but also records the emission positions and intensity gradients of these fluorescence signals in the three-dimensional space of the oil sample. This two-dimensional spatial-spectral information allows analysis to correlate the local concentration and physical location of fluorescent substances, identify non-uniform distributions, local enrichment phenomena, or binding states of dissolved components with specific microstructures, providing direct evidence for determining the potential location of latent faults and the spatial development of the aging process.

[0017] The parameters of suspended impurities obtained from polarization field analysis are fused with the parameters of dissolved components obtained from fluorescence spatial distribution analysis. The comprehensive evaluation index constructed by this scheme simultaneously covers quantitative information and spatial clues of various impurity forms in the oil, including solid, liquid, gas, and dissolved states. This results in a more three-dimensional and comprehensive digital description of the insulation performance and aging state of transformer oil, and its diagnostic dimensions surpass those of conventional reports that simply list various parameters side by side. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the steps of the transformer oil detection method based on polarized light and ultraviolet light described in this invention. Figure 2 The flowchart for establishing a polarization field mapping model and quantitative analysis. Detailed Implementation

[0019] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] This invention relates to a transformer oil detection method based on polarized light and ultraviolet light. The overall implementation scheme of the method is as follows: A polarized light source is used to irradiate a transformer oil sample in a stable flow state, acquiring polarized light field distribution data formed after transmission and scattering. This data contains physical scattering information of suspended particulate matter and moisture in the oil. An ultraviolet light source is used to irradiate the transformer oil sample, exciting specific components to produce characteristic fluorescence. Its fluorescence spectrum and spatial intensity distribution data are acquired, reflecting the chemical composition information of dissolved fault gases and aging products in the oil. A mapping correlation model is established between the polarized light field distribution data and the concentration of suspended particulate matter and trace water content. This model is used to analyze the polarized data to achieve quantitative analysis of physical pollution parameters. A mapping correlation model is also established between the characteristic fluorescence data and the concentration of dissolved fault characteristic gas types and the content of aging product types. This model is used to analyze the fluorescence data to achieve the identification and quantification of chemical components. The analysis results of the above physical and chemical parameters are fused to construct a comprehensive evaluation index for transformer oil condition. Based on this evaluation index and combined with preset oil condition threshold standards, a diagnostic report reflecting the insulation performance and aging degree of the oil sample is finally generated.

[0021] In one embodiment of the present invention, the transformer oil sample to be tested flows at a constant preset flow rate through a detection flow cell made of optical quartz material under the control of a drive pump. The flow channel cross-section of the detection flow cell is rectangular to ensure that the oil flow is in a stable laminar flow state. On one side of the detection flow cell, a semiconductor laser with a wavelength of 650 nanometers is installed as a linearly polarized light source. Its emitted beam becomes linearly polarized light with a purity higher than 100:1 after passing through a Glan Taylor prism, and illuminates the oil sample at an angle perpendicular to the incident surface of the detection flow cell. On the other side of the detection flow cell, i.e., in the direction of the transmitted light path, a first polarization state sensor is arranged to directly receive the transmitted light signal.

[0022] At two specific scattering angles of 30 degrees and 150 degrees relative to the incident light direction, a second polarization state sensor and a third polarization state sensor are installed respectively to receive the scattered light signal. The first, second, and third polarization state sensors are all polarization analysis modules composed of a photodetector, a rotatable waveplate, and a fixed polarizer, capable of simultaneously measuring the complete Stokes parameters of the received light signal. The Stokes parameters are a four-component vector used to fully describe the intensity, degree of polarization, polarization azimuth angle, and ellipticity angle of the light wave. Each polarization state sensor continuously records the dynamic light signal generated by the oil sample flowing through the detection area at a sampling frequency of 1 kHz and outputs a Stokes parameter data stream containing light intensity I, degree of polarization DOP, polarization azimuth angle AoP, and ellipticity angle E.

[0023] In some embodiments, the raw data acquired by the first, second, and third polarization state sensors are transmitted to the central processing unit (CPU). The CPU adds a "Sensor_T" identifier and a high-precision timestamp to the Stokes parameter data stream from the first polarization state sensor, a "Sensor_S30" identifier and a timestamp to the data stream from the second polarization state sensor, and a "Sensor_S150" identifier and a timestamp to the data stream from the third polarization state sensor. A four-dimensional data cube structure is established, with its four dimensions being time t, sensor spatial location identifier s, the four component indices p of the Stokes parameter (e.g., p=1,2,3,4 corresponding to I, DOP, AoP,E respectively), and the parameter value v. The CPU synchronizes the three data streams from the first, second, and third polarization state sensors according to a unified time grid, aligning all data points to the same time series. Subsequently, based on the sensor spatial location identifier s, the synchronized Stokes parameter value v is filled into the corresponding (s,t,p) coordinate position of the data cube. Since the sensor only measures at three discrete spatial angular positions, spatial interpolation processing of the data cube is required to obtain continuous spatial distribution information on the detection cross section.

[0024] Spatial interpolation is understandable; it's a process of estimating the distribution of the entire spatial field based on measurements from a finite number of points. For a data cube at a given moment... and a certain Stokes parameter component Data such as the degree of polarization (DOP) can be reconstructed using an interpolation function at different spatial locations. Optionally, one implementation employs an inverse distance-weighted method for two-dimensional spatial interpolation. The interpolation formula is expressed as: in: Represents coordinates in a spatial plane place, time Stokes parameter components The interpolation result. In this embodiment, the number of sensors is indicated. . Indicates at time ,sensor ( Stokes parameter components measured for “Sensor_T”, “Sensor_S30”, or “Sensor_S150”. The value. It is the weighting coefficient, defined as ,in It is a sensor Spatial location to the interpolation point The Euclidean distance is calculated. Through this process, the measurement data of three discrete spatial points are transformed into polarized light field information with a continuous spatial distribution covering the entire preset detection area. Finally, a four-dimensional dataset integrating time series, spatial coordinates, and complete polarization state information constitutes the polarized light field distribution data for subsequent analysis.

[0025] In one embodiment of the invention, a low-pressure mercury lamp with a center wavelength of 254 nm is used as a pulsed ultraviolet light source, with a pulse width of 10 microseconds and a repetition frequency of 100 Hz. The light beam emitted by the ultraviolet light source is focused by a set of quartz lenses and illuminates a specific detection area of ​​the flow cell. This detection area is adjacent to but does not overlap with the polarized light illumination area to avoid optical crosstalk. The detection area is defined as a circular spot with a diameter of approximately 2 mm. A fiber optic probe is installed at a lateral position 90 degrees perpendicular to the incident direction of the ultraviolet light to collect fluorescence signals. The fiber optic probe is connected to a CCD spectrometer with a resolution better than 0.5 nm. The spectrometer's detection wavelength range covers 300 nm to 600 nm to accommodate the broad-spectrum fluorescence that may be generated in transformer oil. Synchronously with the ultraviolet light pulse, a 512x512 pixel CMOS area array detector is arranged directly above the detection area. This detector is equipped with a bandpass filter to cut off the ultraviolet excitation light, and its frame rate matches the ultraviolet pulse repetition frequency to acquire an image of the fluorescence spatial distribution.

[0026] In some embodiments, the central processing unit controls a GPS-based or high-precision crystal oscillator-based synchronous clock source to add a precise timestamp with the same time reference to each spectral acquisition action of the spectrometer and each image acquisition action of the area array detector, with a timestamp accuracy down to the microsecond level. To achieve spatial registration, a tiny cross-shaped physical marker is etched at the position corresponding to the fiber optic probe on the outer wall of the detection flow cell. This marker is also clearly visible in the fluorescence spatial distribution image acquired by the area array detector. Using the pixel coordinates of this cross-shaped marker in the image as a spatial reference point, a mapping relationship is established between the image pixel coordinate system and the actual physical coordinate system of the detection area. Within each ultraviolet pulse excitation cycle, the spectrometer acquires one frame of fluorescence spectral data, and simultaneously the area array detector acquires a corresponding frame of fluorescence spatial distribution image. The central processing unit associates spectral data frames with the same timestamp with image data frames by comparing the timestamps, forming a "spectrum-image" data pair.

[0027] It is understandable that the raw fluorescence spatial distribution image acquired by the area array detector needs to be processed to obtain accurate quantitative information. Optionally, one implementation is to perform geometric correction on the image to eliminate lens distortion and perform intensity calibration. Intensity calibration is accomplished by imaging a standard fluorescent plate with known absolute intensity under the same optical path and camera settings, establishing a linear relationship between the gray value of each pixel in the image and the absolute fluorescence intensity. The calibration formula is: in: Represents pixel coordinates on the image The absolute fluorescence intensity value at the location. Represents pixel coordinates obtained through calibration. Gain coefficient at that point. Represents pixel coordinates The original grayscale value collected at the location. This represents the same pixel coordinate under completely dark conditions. The background dark noise value was measured at the location. After geometric correction and intensity calibration, each pixel in the image has accurate spatial coordinates and a fluorescence intensity value with physical meaning.

[0028] In some embodiments, the process of constructing a fluorescence spectrum and spatial intensity distribution dataset is systematic. For each "spectrum-image" data pair, based on the spatial registration relationship, the pixel range covered by the area where the fiber optic probe collects fluorescence in the calibrated image is determined. The average absolute fluorescence intensity values ​​of all pixels within this pixel range are calculated, and this average value is bound to the full-frame fluorescence spectrum data acquired by the spectrometer at this time point as the representative spectrum of that spatial location. Simultaneously, the entire calibrated fluorescence spatial distribution image is bound to this timestamp. As the time series progresses, data from all time points are collected, ultimately forming a multidimensional dataset. This multidimensional dataset contains a time dimension t and a two-dimensional spatial coordinate dimension. Or converted physical coordinates And the complete fluorescence spectral intensity distribution corresponding to each time-space point. ,in The wavelength is represented. This dataset, which integrates temporal, spatial, and spectral information, is the fluorescence spectrum and spatial intensity distribution data.

[0029] In one embodiment of the present invention, see [reference] Figure 2In practical implementation, preparing standard transformer oil samples is fundamental to establishing the mapping correlation model. This requires preparing a series of standard transformer oil samples with precisely known suspended particulate matter concentrations and known trace moisture contents. The suspended particulate matter uses standardized ISO intermediate-grade test dust, with strictly defined particle size distribution. Using a precision balance and high-speed dispersion technology, it is uniformly dispersed in the deeply filtered and dried base transformer oil, preparing a sequence of standard samples with suspended particulate matter concentrations of 0 mg / mL, 5 mg / mL, 10 mg / mL, 20 mg / mL, and 50 mg / mL. The trace moisture content is adjusted using a Karl Fischer coulometric moisture analyzer. By quantitatively injecting distilled water into the dried base oil, followed by ultrasonic oscillation and prolonged settling, a sequence of standard samples with trace moisture contents of 5 mg / L, 15 mg / L, 30 mg / L, 50 mg / L, and 80 mg / L is prepared. By orthogonally combining the above particulate matter concentration gradients with moisture content gradients, a composite standard sample library covering multiple concentration points can be prepared.

[0030] In some embodiments, the standard polarization field distribution data of each standard transformer oil sample is acquired under the exact same detection conditions as the oil sample to be tested. Each standard transformer oil sample is injected into the same detection flow cell system as the oil sample to be tested, and passes through the detection zone at the same preset constant flow rate. Irradiation is performed using a 650 nm linearly polarized light source with identical wavelength, power, and polarization state. The first, second, and third polarization state sensors are located at the same spatial angle and collect transmitted and scattered polarized light signals at the same sampling frequency. Data for each standard sample needs to be collected for a sufficiently long time to obtain stable statistical characteristics. The collected raw Stokes parameter data undergoes a complete processing flow: adding high-precision timestamps and spatial location codes to the data streams of each sensor; synchronizing multiple data streams to a unified time grid according to the timestamps; filling the data into a data cube structure according to the location codes; and performing spatial interpolation on the data cube to obtain continuous spatial distribution information. This process ultimately forms the standard polarization field distribution data corresponding to each standard sample.

[0031] It is understandable that extracting a sensitive set of feature parameters from standard polarized light field distribution data is the core of model construction. Analysis revealed a strong correlation between the attenuation of transmitted light polarization degree and the concentration of suspended particulate matter; therefore, the characteristic parameter of transmitted light polarization degree attenuation rate was defined. Its calculation method is as follows in: The polarization degree of the incident light. The polarization degree of the transmitted light after passing through the oil sample. The change in the polarization azimuth angle of the scattered light is sensitive to the birefringence effect caused by moisture. A characteristic parameter is defined for the polarization azimuth angle shift of the scattered light at a specific scattering angle (e.g., 30 degrees). Its calculation method is as follows in: The measured value is for the standard sample. The measured values ​​are for pure, dry base oil. The spatial distribution gradient of the optical field depolarization index reflects non-uniform scattering, and characteristic parameters are defined. The polarization degree two-dimensional distribution map is obtained by calculating the magnitudes of the partial derivatives in two spatial directions of the interpolated polarization degree two-dimensional distribution map. These feature parameters are extracted from each standard sample to form a feature parameter set. In some embodiments, the mapping correlation model is established using the support vector regression method in machine learning algorithms. The feature parameter set extracted from the standard polarization light field distribution data of the standard samples is used as the input matrix. Each row of the matrix represents a standard sample, and each column represents a feature parameter (e.g., ...). (The known concentration of suspended particulate matter in the standard sample). and trace moisture content As the target output, the entire standard sample dataset is divided into training and validation sets. A multi-output support vector regression model is trained using the training set data, which learns from a multi-dimensional feature space to a two-dimensional concentration space. The model exhibits a complex nonlinear mapping relationship. Its performance is evaluated and optimized using the root mean square error between predicted and actual concentrations on the validation set.

[0032] Optionally, after obtaining the trained support vector regression model, the transformer oil sample to be tested is analyzed. Under the same detection conditions, the oil sample is illuminated by a 650 nm linearly polarized light source, and Stokes parameters are collected by polarization state sensors positioned along the transmission direction and at 30° and 150° scattering angles. The polarization field distribution data, obtained through time synchronization, spatial encoding, and interpolation, undergoes the same feature parameter extraction process to obtain the feature parameter vector of the oil sample. This feature parameter vector is input into the established support vector regression mapping association model. The model directly outputs two values: the quantitative analysis results of the suspended particulate matter concentration and the quantitative analysis results of the trace water content parameter in the transformer oil sample.

[0033] In one embodiment of the present invention, the preparation of standard transformer oil samples requires covering common soluble fault characteristic gases and aging products in transformer oil. The standard samples are formulated from base transformer oil that has undergone rigorous purification. Soluble fault characteristic gases include hydrogen (H2), carbon monoxide (CO), methane (CH4), ethane (C2H6), ethylene (C2H4), and acetylene (C2H2). A precise gas distribution device is used to dissolve high-purity single gases or mixtures of gases in specific proportions into the base oil, preparing standard samples with low, medium, and high concentration gradients for each gas, covering the typical fault diagnosis thresholds. Aging products mainly target furan compounds such as furfural (2-furancarbaldehyde) and low-molecular-weight organic acids. Analytical-grade chemicals are dissolved in the base oil using a gravimetric method to prepare standard solutions with varying concentration gradients. The final standard sample library contains a series of standard transformer oil samples containing known single components or mixtures of components in known proportions, each sample being precisely labeled with the type and concentration information of its components.

[0034] In some embodiments, the acquisition of standard fluorescence spectra and spatial intensity distribution data for each standard transformer oil sample strictly follows the detection conditions of excitation with a pulsed ultraviolet light source with a center wavelength of 254 nm, acquisition of fluorescence spectral signals using a spectrometer, and simultaneous acquisition of fluorescence spatial distribution images using an area array detector. Each standard sample is sequentially injected into the detection system, and the sample is excited using a pulsed ultraviolet light source with a center wavelength of 254 nm at the same power and repetition frequency. The spectrometer and area array detector acquire data synchronously. The spectrometer records the fluorescence spectrum in the wavelength range of 300 nm to 600 nm, and the area array detector records the two-dimensional spatial distribution image of fluorescence in the detection area. The acquired raw time-series spectral data and image data are processed through the same time synchronization, spatial registration, geometric correction, and intensity calibration process as described above under the detection conditions of excitation with an ultraviolet light source and simultaneous acquisition by the spectrometer and area array detector, ultimately forming standard fluorescence spectra and spatial intensity distribution data for each standard sample, including both spectral and spatial dimensions.

[0035] It is understandable that feature extraction from standard fluorescence spectra and spatial intensity distribution data aims to obtain unique identifiers for each component. From the time-averaged spectrum of each standard sample, characteristic fluorescence spectral fingerprints are extracted, including the positions of characteristic peaks. The wavelength corresponding to the local intensity maximum on the spectral curve; the full width at half maximum (FWHM), the peak width at half the height of the characteristic peak; and the peak area ratio. This refers to the area ratio between different characteristic peaks. Characteristic fluorescence spatial distribution patterns are extracted from the average fluorescence spatial distribution image of each standard sample. These patterns include the fluorescence intensity spatial uniformity index. The fluorescence diffusion profile parameter is obtained by calculating the coefficient of variation of the intensity values ​​of all pixels in the image; and the fluorescence diffusion profile parameter is obtained by fitting a function to the fluorescence intensity decay curve from the center of the excitation spot outwards. Different components, due to differences in their molecular structure and diffusion characteristics, will exhibit different combinations of characteristic fluorescence spectral fingerprints and characteristic fluorescence spatial distribution patterns. For a clearer illustration, see Table 1 for the characteristic fluorescence spectral fingerprints of some typical components: Table 1: Characteristic Fluorescence Spectral Fingerprint Table of Typical Failure Gases and Aging Products In some embodiments, a step-by-step strategy is adopted to establish a mapping correlation model for resolving the contributions of each component from mixed fluorescence data. A spectral unmixing algorithm is used to process the fluorescence spectral data. For a mixed spectral data matrix X (rows are wavelengths, columns are different measurement points or time points) containing k known components, it is assumed to be a linear mixture of the pure component spectral matrix S (rows are wavelengths, columns are components) and the contribution coefficient matrix C (rows are components, columns are measurement points), and is affected by the noise matrix E. The model is expressed as: in: It is the measured matrix of mixed fluorescence spectral data. It is the characteristic fluorescence spectral fingerprint matrix of the pure component, either to be determined or a known reference. It is the contribution coefficient matrix corresponding to each component. The residual matrix is ​​given. Using algorithms such as nonnegative matrix factorization, under the constraint of some known prior spectral information (as shown in Table 1), S and C can be iteratively solved from X. The element values ​​in the contribution coefficient matrix C are proportional to the concentration of the corresponding component. Next, a pattern recognition algorithm is used to process the fluorescence spatial distribution data. The extracted fluorescence intensity spatial uniformity index H_uniform and fluorescence diffusion profile parameters are matched with patterns in the standard sample library. The similarity score between the test data and the characteristic fluorescence spatial distribution patterns of each standard component is calculated, and the similarity score is proportional to the component content.

[0036] Optionally, after model training and optimization, the transformer oil sample to be tested is identified and quantified. The sample is irradiated with a pulsed ultraviolet light source with a center wavelength of 254 nm. A spectrometer is used to collect fluorescence spectral signals, and a regional array detector is used to acquire spatial distribution images of the fluorescence. The acquired time-series spectral data and image data are synchronized in time and spatially registered. The obtained fluorescence spectral and spatial intensity distribution data are input into the established spectral unmixing and pattern recognition mapping association model. The spectral unmixing part of the model outputs the contribution of characteristic fluorescence spectral fingerprints of each potential component, and the pattern recognition part outputs the similarity to the characteristic fluorescence spatial distribution patterns of each component. Combining the contribution and similarity information and comparing it with a standard database, the final output is the identification and quantification results of the type and concentration of dissolved fault characteristic gases and the types and contents of aging products in the oil.

[0037] In one embodiment of the present invention, in a specific implementation, the data fusion process converts the quantitative analysis results of suspended particulate matter concentration, the quantitative analysis results of trace water content parameters, and the identification and quantification results of the types and concentrations of dissolved fault characteristic gases and the types and contents of aging products into a unified standardized dimensional space. Each parameter is converted into a scale value in the range of zero to one through a linear normalization method. Within the standardized dimensional space, each standardized parameter is assigned a preset weight coefficient according to its importance in the transformer oil condition assessment. For example, the weight coefficient for suspended particulate matter concentration is set to 0.15, the weight coefficient for trace water content parameters is set to 0.20, the weight coefficient for acetylene concentration is set to 0.25, and the weight coefficient for furfural content is set to 0.20. The remaining weights for other fault gases and aging products are assigned according to their typical effects. Based on the standardized values ​​of all parameters and their corresponding weight coefficients, a weighted fusion algorithm is used to calculate the preliminary score of the transformer oil condition. The weighted fusion calculation formula is expressed as follows: in: This represents a preliminary score for the condition of the transformer oil. , , , These represent the concentration of suspended particulate matter, the trace water content parameter, and the first... The concentration of the faulty gas, the first The weighting coefficients corresponding to the content of various aging products. , , , This represents the standardized value of the corresponding parameter. and These represent the number of identified fault gas types and the number of aging product types, respectively. The spatial uniformity index of the polarized light field distribution data and the fluorescence diffusion profile index of the fluorescence spatial distribution data are introduced as correction factors. The spatial uniformity index of the light field is obtained by calculating the variance of the transmitted light intensity on the detection cross section, and the fluorescence diffusion profile index is obtained by fitting the fluorescence intensity decay curve to obtain the exponential decay constant. Based on the pre-established empirical relationship between these indices and the stability of the transformer oil condition, the measured values ​​are converted into a correction coefficient. This correction coefficient is used to perform a product correction on the preliminary transformer oil condition score, resulting in the final comprehensive evaluation index of the transformer oil condition.

[0038] In some embodiments, multiple transformer oil condition levels are preset and associated with threshold ranges and descriptive text. For example, the condition level "Excellent" corresponds to a comprehensive evaluation index range of 0 to 0.2, "Good" corresponds to a range of 0.2 to 0.4, "Medium" corresponds to a range of 0.4 to 0.6, "Poor" corresponds to a range of 0.6 to 0.8, and "Severe" corresponds to a range of 0.8 to 1.0. Each level is associated with preset insulation performance descriptions and aging degree descriptions. The comprehensive evaluation index of the transformer oil condition of the transformer oil sample to be tested is compared with these threshold ranges to determine its condition level. The insulation performance description and aging degree description corresponding to that level are extracted as the basic content of the diagnostic report. The type tendency information of the main pollutants is extracted from the polarization field distribution data analysis results, such as metal particle dominance or fiber contamination signs. The type tendency information of the main fault gases or aging products is extracted from the fluorescence spectrum and spatial intensity distribution data analysis results, such as ethylene dominance overheating faults or furan compounds indicating aging. This tendency information, along with specific quantitative values ​​such as suspended particulate matter concentration, trace water content, fault gas concentration, and aging product content, is filled into the basic content of the diagnostic report to generate a detailed diagnostic report containing condition level, key parameter values, and material tendency analysis.

[0039] It is understandable that the process of converting the light field spatial uniformity index and the fluorescence diffusion profile index into correction coefficients is based on empirical relationships. When the light field spatial uniformity index is below a preset threshold and the fluorescence diffusion profile index shows rapid decay, the correction coefficient approaches 1, indicating that the oil condition is stable and the initial score does not require significant adjustment. When the light field spatial uniformity index is high or the fluorescence diffusion profile index decays slowly, the correction coefficient deviates from 1, adjusting the initial score upwards or downwards. This dynamic correction mechanism ensures that the comprehensive evaluation index of transformer oil condition not only reflects the absolute concentration of contaminants and dissolved components but also incorporates information on their uniformity and stability of distribution in the oil. Optionally, the output format of the detailed diagnostic report includes a text summary and a structured data table. The text summary summarizes the condition level and main conclusions, while the structured data table lists all quantitative parameters, standardized values, weighting coefficients, and intermediate calculation results for further analysis and verification.

[0040] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for detecting transformer oil based on polarized light and ultraviolet light, characterized in that, include: A polarized light source is used to irradiate a transformer oil sample under stable flow to obtain polarized light field distribution data formed by the transmission and scattering of the transformer oil sample. The transformer oil sample to be tested is irradiated with an ultraviolet light source to excite specific components in the transformer oil sample to produce characteristic fluorescence, and the fluorescence spectrum and spatial intensity distribution data are obtained. A mapping correlation model is established between polarized light field distribution data and suspended particulate matter and water content in oil. Based on the mapping correlation model, the polarized light field distribution data is analyzed to quantitatively resolve the parameters of suspended particulate matter concentration and trace water content in oil. A mapping correlation model is established between characteristic fluorescence spectra and characteristic gases of solubility failures and aging products in oil. Based on the mapping correlation model, the fluorescence spectra and spatial intensity distribution data are analyzed to identify and quantify the types and concentrations of characteristic gases of solubility failures and the types and contents of aging products in oil. The parameters of suspended particulate matter concentration and trace water content in the oil are fused with the types and concentrations of dissolved fault characteristic gases and the types and contents of aging products to construct a comprehensive evaluation index for transformer oil condition. Based on the comprehensive evaluation index of transformer oil condition and combined with the preset transformer oil condition threshold standard, a diagnostic report on the insulation performance and aging degree of the transformer oil sample to be tested is generated.

2. The transformer oil detection method based on polarized light and ultraviolet light according to claim 1, characterized in that, The process of irradiating a transformer oil sample in a stable flow state with a polarized light source to obtain polarized light field distribution data formed by the transmission and scattering of the transformer oil sample specifically includes: The transformer oil sample to be tested is passed through an optically transparent detection flow cell at a preset constant flow rate. A linearly polarized light source of a single wavelength or narrow band is used to irradiate the transformer oil sample to be tested in the detection flow cell at a preset incident angle. Multiple polarization state sensors are arranged at multiple preset spatial angle positions in the detection flow cell to collect polarized light signals in the transmission direction and scattered polarized light signals at at least two specific scattering angles, respectively. Each of the polarization state sensors synchronously records the Stokes parameters of the received optical signal, which include light intensity, degree of polarization, polarization azimuth angle, and ellipticity angle. Stokes parameters collected from different spatial angles are aligned and integrated according to timestamps and spatial locations to form polarization field distribution data that describes the spatiotemporal evolution of the polarization field of the transformer oil sample under test during the detection period.

3. The transformer oil detection method based on polarized light and ultraviolet light according to claim 2, characterized in that, The process of irradiating the transformer oil sample with an ultraviolet light source to excite specific components in the sample to produce characteristic fluorescence and obtaining fluorescence spectrum and spatial intensity distribution data specifically includes: A pulsed or continuous ultraviolet light source is used, wherein the center wavelength of the ultraviolet light source is located in the band that can excite the aromatic hydrocarbon compounds and fault characteristic gas derivatives in the transformer oil to produce fluorescence; The ultraviolet light source beam is focused onto the detection area in the detection flow cell that overlaps with or is adjacent to the polarized light irradiation area. The excited fluorescence signal is collected by using a spectrometer in the detection area at a certain angle to the incident light, and the resolution of the spectrometer is sufficient to distinguish the fluorescence characteristic peaks of different components. Simultaneously, an area array detector is used to acquire images of the spatial distribution of fluorescence generated in the detection area under ultraviolet light irradiation; The time-series fluorescence spectral data acquired by the spectrometer and the time-series fluorescence spatial distribution image acquired by the area array detector are synchronized in time and registered in space to obtain fluorescence spectral and spatial intensity distribution data containing spectral and two-dimensional spatial information.

4. The transformer oil detection method based on polarized light and ultraviolet light according to claim 3, characterized in that, The process involves establishing a mapping correlation model between polarized light field distribution data and the content of suspended particulate matter and water in oil. Based on this model, the polarized light field distribution data is analyzed to quantitatively determine the concentration of suspended particulate matter and trace water content parameters in the oil. Specifically, this includes: A series of standard transformer oil samples with known gradient concentrations of suspended particulate matter and known trace water content were prepared. Using the same detection conditions as those used to acquire the polarization field distribution data of the transformer oil sample to be tested, standard polarization field distribution data of each of the standard transformer oil samples was acquired. The variation law of Stokes parameter in the standard polarized light field distribution data was analyzed, and a set of characteristic parameters sensitive to the changes in suspended particulate matter concentration and moisture content was extracted. The set of characteristic parameters includes the attenuation rate of polarization degree of transmitted light, the polarization azimuth angle offset of scattered light at a specific scattering angle, and the spatial distribution gradient of the light field depolarization index. Based on the set of feature parameters and the known concentration of suspended particulate matter and trace water content, a mapping and correlation model from the set of feature parameters to the concentration of suspended particulate matter and trace water content is established by training through multivariate nonlinear regression or machine learning algorithms. The polarization field distribution data of the transformer oil sample to be tested is input into the mapping correlation model to calculate the quantitative analysis results of the suspended particulate matter concentration and trace water content parameters in the transformer oil sample to be tested.

5. The transformer oil detection method based on polarized light and ultraviolet light according to claim 4, characterized in that, The process involves establishing a mapping correlation model between characteristic fluorescence spectra and characteristic gases and aging products of dissolved failures in oil. Based on this model, the fluorescence spectra and spatial intensity distribution data are analyzed to identify and quantify the types and concentrations of characteristic gases of dissolved failures and the types and contents of aging products in the oil. Specifically, this includes: A series of standard transformer oil samples containing single or mixed known types and concentrations of soluble fault characteristic gases and aging products were prepared. Using the same detection conditions as those used to obtain the fluorescence spectrum and spatial intensity distribution data of the transformer oil sample to be tested, standard fluorescence spectrum and spatial intensity distribution data of each of the standard transformer oil samples were obtained. Feature extraction is performed on the standard fluorescence spectrum and spatial intensity distribution data to obtain the characteristic fluorescence spectral fingerprint and characteristic fluorescence spatial distribution pattern of each known component. The characteristic fluorescence spectral fingerprint includes the characteristic peak position, full width at half maximum (FWHM), and peak area ratio. The characteristic fluorescence spatial distribution pattern includes the fluorescence intensity spatial uniformity index and diffusion profile. Using spectral unmixing and pattern recognition algorithms, a mapping and correlation model is established to extract the contribution of characteristic fluorescence spectral fingerprints and the similarity of characteristic fluorescence spatial distribution patterns of each component from the fluorescence spectra and spatial intensity distribution data to be tested. The contribution and similarity are proportional to the component concentration and content. The fluorescence spectrum and spatial intensity distribution data of the transformer oil sample to be tested are input into the mapping correlation model, and the identification and quantification results of the type and concentration of dissolved fault characteristic gases and the types and contents of aging products in the oil are obtained by solving the model.

6. The transformer oil detection method based on polarized light and ultraviolet light according to claim 5, characterized in that, The concentration of suspended particulate matter and trace water content in the oil are fused with the types and concentrations of dissolved fault characteristic gases and the types and contents of aging products to construct a comprehensive evaluation index for transformer oil condition, specifically including: The quantitative analysis results of the suspended particulate matter concentration, the quantitative analysis results of the trace water content parameters, and the identification and quantification results of the type and concentration of the characteristic gas of the dissolved failure and the type and content of the aging products are uniformly converted into a preset standardized dimension space. Within the standardized dimension space, each parameter is assigned a weighting coefficient based on its importance in the assessment of transformer oil insulation performance and aging status. Based on the standardized value of each parameter and its corresponding weight coefficient, a weighted fusion algorithm is used to calculate a comprehensive preliminary score of transformer oil condition. The optical field spatial uniformity index in the polarization optical field distribution data and the fluorescence diffusion profile index in the fluorescence spatial distribution data are introduced as correction factors for the preliminary score of transformer oil condition, and the preliminary score is dynamically corrected. The dynamically corrected score will be used as the comprehensive evaluation index for transformer oil condition.

7. The transformer oil detection method based on polarized light and ultraviolet light according to claim 6, characterized in that, Based on the comprehensive evaluation index of transformer oil condition, and combined with the preset transformer oil condition threshold standard, a diagnostic report on the insulation performance and aging degree of the transformer oil sample to be tested is generated, specifically including: Multiple transformer oil condition levels are preset, each level corresponds to a threshold range of a comprehensive evaluation index for transformer oil condition, and is associated with different insulation performance descriptions and aging degree descriptions; The transformer oil condition comprehensive evaluation index of the transformer oil sample to be tested is compared with the threshold range of the transformer oil condition comprehensive evaluation index to determine the condition level of the transformer oil sample to be tested. Extract the preset insulation performance description and aging degree description belonging to the aforementioned state level as the basis for the diagnostic report; The analysis results of the polarization field distribution data are used to extract the type tendency information of the main pollutants, and the analysis results of the fluorescence spectrum and spatial intensity distribution data are used to extract the type tendency information of the main failure gases or aging products. The type and tendency information of the main pollutants and the type and tendency information of the main fault gases or aging products, combined with the specific values ​​in the quantitative analysis results and identification and quantification results, are filled into the basic content of the diagnostic report to generate a detailed diagnostic report that includes the state level, key parameter values ​​and substance tendency analysis.

8. The transformer oil detection method based on polarized light and ultraviolet light according to claim 7, characterized in that, The process of aligning and integrating Stokes parameters collected from different spatial angles according to timestamps and spatial locations to form polarization field distribution data describing the spatiotemporal evolution of the polarization field of the transformer oil sample under test during the detection period specifically includes: Add high-precision timestamps and sensor spatial location codes to the Stokes parameter data streams acquired by each polarization state sensor; Establish a data cube structure with time as the horizontal axis and spatial location and Stokes parameters as the dimensions; Synchronize the Stokes parameter sequences from all polarization state sensors to a unified time grid according to the timestamps; Based on the sensor spatial location encoding, the synchronized Stokes parameters are filled into the corresponding positions of the data cube structure; Spatial interpolation is performed on the data cube structure to obtain continuously distributed polarized light field information at a preset spatial resolution, forming complete polarized light field distribution data.

9. The transformer oil detection method based on polarized light and ultraviolet light according to claim 8, characterized in that, The time-series fluorescence spectral data acquired by the spectrometer and the time-series fluorescence spatial distribution image acquired by the area array detector are time-synchronized and spatially registered to obtain fluorescence spectral and spatial intensity distribution data containing spectral and two-dimensional spatial information, specifically including: Add a timestamp based on the same time reference to each spectral data frame acquired by the spectrometer and each image data frame acquired by the area array detector. The physical position of the optical fiber probe of the spectrometer in the image of the array detector is used as the reference point for spatial registration. Based on timestamp alignment, the spectral data and image data corresponding to each time point are associated to form a "spectrum-image" data pair with time sequence. Geometric correction and intensity calibration are performed on the fluorescence spatial distribution image acquired by the area array detector to ensure that the spatial coordinates and fluorescence intensity values ​​of each pixel in the image are accurate. The spatial coordinates and intensity information of each pixel in the calibrated image are associated and extended with the spectral data of its corresponding time point to construct a dataset containing time, two-dimensional spatial coordinates, and fluorescence spectral intensity distribution at the corresponding spatial location point, namely fluorescence spectrum and spatial intensity distribution data.

10. The transformer oil detection method based on polarized light and ultraviolet light according to claim 9, characterized in that, The spatial uniformity index of the polarization field distribution data and the fluorescence diffusion profile index of the fluorescence spatial distribution data are introduced as correction factors for the preliminary score of transformer oil condition, and the preliminary score is dynamically corrected, specifically including: The variance of transmitted light intensity or degree of polarization on the detection section is calculated from the polarized light field distribution data and used as an index of the spatial uniformity of the light field to characterize the uniformity of the oil. The fluorescence intensity decay curve characteristic parameter from the excitation center outward is extracted from the fluorescence spatial distribution data and used as a fluorescence diffusion profile index to characterize the diffusion or distribution of dissolved substances. An empirical relationship is established between the light field spatial uniformity index and the fluorescence diffusion profile index and the stability of the oil condition. When the light field spatial uniformity index is below the threshold and the fluorescence diffusion profile index shows rapid decay, the oil condition is determined to be relatively stable. Based on the aforementioned empirical relationship, the optical field spatial uniformity index and the fluorescence diffusion profile index are converted into correction coefficients for the preliminary score of the transformer oil condition. The initial transformer oil condition score is corrected by multiplying the correction coefficient to obtain the final comprehensive evaluation index of transformer oil condition calibrated by stability factors.

Images