A transformer insulation state online monitoring method based on fiber vibration imaging

Abstract

The application provides a transformer insulation state online monitoring method based on optical fiber vibration imaging, and comprises the following steps: obtaining a vibration signal of a detection part of a transformer by using a distributed optical fiber; converting the vibration signal into an optical fiber vibration original image based on a principle; pre-processing the optical fiber vibration original image; extracting typical features of the pre-processed optical fiber vibration original image; inputting the extracted typical features into a constructed transformer insulation evaluation model; and solving the typical features by using the transformer insulation evaluation model to obtain transformer insulation state data. The application utilizes the characteristics of transformer vibration parameter changes caused by insulation material aging, obtains vibration parameter features based on optical fiber vibration imaging, and evaluates the transformer insulation state, so that the insulation state can be evaluated under the condition of no sampling or power-off maintenance.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of transformers, and particularly relates to a transformer insulation state online monitoring method based on optical fiber vibration imaging. BACKGROUND

[0002] During the operation of a transformer, the insulation paper is subjected to long-term action of external stress, and a series of physical changes and chemical changes are generated, so that the electrical performance and mechanical performance of the insulation paper gradually decrease, which is called insulation aging of the insulation paper. The forms of insulation aging of the insulation paper mainly include thermal aging, electrical aging and mechanical aging. Thermal aging is the main form of insulation aging of the insulation paper. The main component of the insulation paper is cellulose, which will undergo thermal degradation under the action of high temperature, so that the main chain of macromolecules is broken to generate small molecular products, and the molecular weight is reduced. Due to the change of the microstructure of the insulation material, the mechanical performance is degraded. The transformer is excited by electromagnetic force to generate mechanical vibration during the operation, and the insulation paper generates micro defects under the mechanical stress and gradually expands, which causes material damage, which is called mechanical aging of the insulation paper. In recent years, some scholars have studied the mechanical performance degradation of the insulation paper under the combined action of machine and heat. The promotion effect of the combined aging of machine and heat on the mechanical performance degradation of the insulation paper is obviously greater than that of thermal aging, and the more intense the mechanical vibration is, the more significant the decrease of the polymerization degree and the tensile strength of the insulation paper is. In recent years, some scholars have begun to pay attention to the influence of mechanical aging on the mechanical performance of the insulation paper when the thermal aging of the insulation paper is considered. Therefore, the above researches are difficult to reflect the change rule of the mechanical performance of the insulation paper under the combined action of heat, electricity and machine. There is a lack of necessary support for the research on the change of the vibration characteristics of the transformer caused by the aging of the insulation material. At present, the following methods are used for the evaluation of the insulation state of the transformer: (1) chemical analysis method for measuring the content of dissolved gas in oil

[0003] From the perspective of chemical analysis, the insulation material will generate CO, CO2 and trace amounts of hydrocarbons when it is subjected to thermal degradation. By measuring the content of various gases dissolved in the insulation oil, the aging condition of the insulation material can be estimated. Kan et al. studied the relationship between the content of CO and CO2 in the transformer oil and the average polymerization degree of the insulation paper, and found that when the content of CO and CO2 is 1 mL / g, the residual polymerization degree of the insulation paper is 50%; when the content is 3 mL / g, the residual polymerization degree is 30%

[24] However, this method has certain defects: in the transformer, in addition to carbon monoxide and carbon dioxide generated by the aging of the insulation paper, the degradation of the transformer oil will also generate these two gases. In addition, the production of carbon monoxide and carbon dioxide is related to the structure, model and capacity of the transformer. Therefore, it has great limitations to determine the service life of the insulation material in different types of transformers according to this method.

[0004] (2) oil-paper insulation frequency domain dielectric response analysis

[0005] From the electrical analysis point of view, the insulating material of transformer gradually loses water and forms aging derivatives (Cu2S, acids) during the aging process. The frequency domain dielectric response (such as dielectric loss factor, etc.) can reflect the aging degree of the insulating material. Liao Ruojin et al. carried out a large number of test work on the complex capacitance C * and dielectric loss factor tanδ of insulating paper (paperboard) with different aging degrees and different water contents. The test results show that the aging significantly increases the low-frequency part (less than 0.1 Hz) of the complex capacitance C * and dielectric loss factor tanδ, while the high-frequency part changes little [25,26] . Liu Ji et al. improved the method based on the transformer XY equivalent model and the composite dielectric dielectric model by using the frequency band optimization calculation method. In addition, the failure of the insulating material will also cause the change of the insulation resistance of the power transformer

[27] . This method has high sensitivity for judging the insulation state of the transformer, and the insulation resistance table is small in size, easy to carry, and convenient to test. However, the detection of the insulation resistance is easily affected by temperature, and there is an electrical connection with the transformer during measurement, which can only be carried out when the equipment is shut down, and cannot be detected online.

[0006] (3) Determining tensile strength and polymerization degree

[0007] From the point of view of material mechanics, the most intuitive manifestation of the aging of transformer insulating materials is the change of the mechanical strength of the material. It is pointed out in some documents that the aging degree of insulating paper mainly depends on its mechanical strength

[28] . At the same time, the decrease of the polymerization degree of insulating paper effectively reflects the decrease of its mechanical strength 29 . Hill et al. found that the tensile strength of insulating paper decreases with the increase of aging time and the rise of aging temperature, and the aging degree of insulating paper can be estimated by the tensile strength 30 . Hill et al. also studied the relationship between the polymerization degree and the tensile strength of insulating paper, and found that the change trends of the two are basically the same. As mentioned earlier, when the tensile strength or polymerization degree of insulating paper decreases to 25% of the initial value, the material is considered to be failed and the life is terminated.

[0008] The existing detection methods are difficult to accurately evaluate the insulation state of the transformer under the premise of not stopping sampling and de-energizing maintenance. In addition, there are many types of transformers, and the differences in design and manufacture, operation environment, and operation and maintenance conditions are great, and the implementation of the above methods has certain limitations. SUMMARY

[0009] This invention solves the problem that existing detection methods are difficult to accurately assess the insulation status of transformers without stopping the power supply for sampling or power outage maintenance. It proposes an online monitoring method for transformer insulation status based on fiber optic vibration imaging. By utilizing the characteristics of changes in transformer vibration parameters caused by the aging of insulation materials, the method acquires vibration parameter features based on fiber optic vibration imaging to assess the insulation status of the transformer. This method can achieve insulation status assessment even without stopping the power supply for sampling or power outage maintenance.

[0010] To achieve the above objectives, the following technical solution is proposed:

[0011] A method for online monitoring of transformer insulation condition based on fiber optic vibration imaging includes the following steps:

[0012] S1, using distributed optical fiber to acquire vibration signals from the detection points of the transformer;

[0013] S2, based on - The OTDR principle converts vibration signals into raw images of fiber optic vibration;

[0014] S3, preprocess the original image of fiber vibration;

[0015] S4, Extract typical features from the preprocessed original image of fiber vibration;

[0016] S5. Input the extracted typical features into the constructed transformer insulation evaluation model;

[0017] S6, the transformer insulation assessment model solves for typical characteristics to obtain transformer insulation status data.

[0018] This invention obtains typical characteristic information of transformers through fiber optic vibration imaging, and obtains the insulation status of the transformers by analyzing the typical characteristic information, realizing condition monitoring of key power equipment from a completely new perspective of intuitive imaging. Because it uses distributed optical fibers to acquire vibration signals from the detection points of the transformer, this invention can assess the insulation status even under conditions of non-stop sampling or power outage maintenance.

[0019] Preferably, S1 specifically includes the following steps: using distributed optical fibers to wrap around the transformer body to directly collect vibration signals from the transformer core and windings.

[0020] Preferably, S2 specifically includes the following steps: based on - The Rayleigh backscatter amplitude detection technology of OTDR extracts the vibration signal at any point in the global coordinate system, and draws the original image of fiber vibration based on the corresponding vibration signal in the global coordinate system.

[0021] Preferably, the preprocessing in S3 includes: grayscale conversion, image enhancement, and image segmentation.

[0022] As preferred, the S4 specifically comprises the following steps: extracting typical features of the pre-processed optical fiber vibration original image based on a gray level co-occurrence matrix method, the typical features including a texture contrast, an optical fiber vibration image energy density, a local texture variable, a texture complexity and a graph correlation parameter.

[0023] As preferred, the transformer insulation evaluation model is:

[0024]

[0025] wherein Y(x, t) is a state quantity of the monitoring image at the t time, Y R (x, t0) is a state quantity at an initial time of insulation, x is a variable of the typical feature, and Omega represents an imaging space of the optical fiber image, and adjusting the range thereof can detect a local or global of the transformer body.

[0026] As preferred, the variable x of the typical feature includes:

[0027] a texture contrast:

[0028]

[0029] an optical fiber vibration image energy density:

[0030]

[0031] a local texture variable:

[0032]

[0033] a texture complexity:

[0034]

[0035] a graph correlation parameter:

[0036]

[0037] wherein p is a gray level co-occurrence matrix element, n is a pixel number, S i,j is a pixel area.

[0038]

[0039] The beneficial effects of the present application are: the present application obtains the typical feature information of the transformer through fiber vibration imaging, obtains the insulation state of the transformer through the analysis of the typical feature information, and realizes the state monitoring of the key power equipment from the new visual angle of intuitive imaging. Since the distributed optical fiber is used to obtain the vibration signals of the detection positions of the transformer, the insulation state can be evaluated under the condition of no sampling or power-off maintenance. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 is a flowchart of the embodiment method;

[0041] Figure 2 is an embodiment phase demodulation principle schematic diagram;

[0042] Figure 3 is an embodiment fiber vibration image after preprocessing. DETAILED DESCRIPTION

[0043] Embodiment:

[0044] The present embodiment proposes a transformer insulation state online monitoring method based on fiber vibration imaging, referring to Figure 1 , comprising the following steps:

[0045] S1, using a distributed optical fiber to obtain the vibration signals of the detection positions of the transformer; the distributed optical fiber is wrapped around the body of the transformer to directly collect the vibration signals of the core and winding of the transformer.

[0046] The transformer body refers to the core and winding components that realize high-low voltage conversion. In addition to the body, the power transformer also includes an oil tank, a cooling device, a voltage regulating device, a protection device, etc. In addition to the insulating material, the aging of the metal components of the transformer is not obvious. The insulating material is mainly used for the current-carrying winding and its tap to avoid turn-to-turn short circuit and ground discharge. The copper conductor is wrapped with insulating paper and wound into a wire cake according to certain process requirements, and is stacked into high-low voltage windings by spacers. The mass, spring and damping units are commonly used to describe the winding vibration model, wherein the spring part is equivalent to the stiffness of the insulating paper and the insulating spacer. In the normal state where the metal conductor does not age obviously, the overall stiffness of the winding assembly depends on the stiffness of the series insulating material. The change of the assembly stiffness caused by the aging of the insulating material can be obtained through the inversion of the vibration signal.

[0047] S2, based on -OTDR principle converts the vibration signal into a fiber vibration original image; based on -Rayleigh backscattering light amplitude detection technology of OTDR extracts the vibration signal of any point under the global coordinates, and draws the fiber vibration original image according to the corresponding vibration signal of the global coordinates.

[0048] -OTDR as a kind of distributed optical fiber sensing technology, by detecting the intensity, frequency, phase, polarization state of scattered light changes along the optical fiber axial laying environment to perceive the changes in state. At present, based on -OTDR Rayleigh backscattering light amplitude detection technology in the vibration monitoring is the most mature. However, the technology can only achieve the position and frequency of the detection of vibration, even when the external vibration is too large will lead to the harmonic effect of the amplitude, the quantitative detection of strain is very disadvantageous. Research found that the phase change of the scattered light and strain are proportional. Therefore, using phase demodulation algorithm to obtain the phase information of each scattering point on the optical fiber, has the potential of quantitative detection.

[0049] The length of the optical fiber is L, and a plurality of scattering points are randomly distributed in the optical fiber. When the optical pulse propagates forward (in the direction of the yellow arrow), backscattering occurs at each scattering point. For the sake of intuition, it is represented as N discrete reflectors, and each reflector can be regarded as the result of the joint action of M scattering points randomly distributed in a specific length of optical fiber. Since the amplitude and phase of the backscattering signal of each scattering point are also random, the sum of the backscattering signals of the M scattering points can also be regarded as random in the complex plane, which can be represented as:

[0050]

[0051] In the above formula, k represents the Kth reflector, i represents the ith scattering point in a specific length, a i , is the phase and amplitude of the ith scattering point, A k , φ k is the phase and amplitude of the kth reflector, and the influence of the polarization direction is ignored. When ΔL is the length of the optical pulse, i.e. equal to the optical pulse interference length, the APD detects the photocurrent size as:

[0052]

[0053] When a certain place on the optical fiber is affected by external factors such as temperature, vibration, etc., the core radius, optical fiber length, refractive index and other parameters at that position will change, thereby changing the phase size, ultimately affecting the photocurrent size detected by the APD, realizing distributed optical fiber sensing.

[0054] Since -OTDR is a distributed continuous measurement, realizing the vibration measurement and imaging of transformer internal windings and other components, the primary problem to be solved is to accurately extract the vibration signal of any point in the global coordinate, i.e. the positioning accuracy of distributed sensing. Similar to spatial resolution, positioning information is also affected by the coherence of each scattering point within the light pulse width. From the formula: It is known that the positioning accuracy is limited by the optical pulse width. In the formula, w is the optical pulse width, n is the fiber refractive index, c is the speed of light, and Δz is the spatial resolution. Assuming the fiber refractive index n equals 1.5, then when the system's optical pulse is 100 ns, the theoretical spatial resolution of the system is 10 m. The actual resolution is also affected by factors such as the frequency drift of the light source, the system's signal-to-noise ratio, and the amplifier's self-gain interference. How to further improve the system's spatial resolution under the premise of a fixed optical pulse width through noise suppression (temperature drift) and winding design (process) is one of the research contents of this section.

[0055] Traditional The OTDR system architecture achieves distributed sensing by directly detecting the Rayleigh backscattered signal in the optical fiber. However, the backscattered signal in single-mode fiber is extremely weak, especially at the far end of the fiber where the signal-to-noise ratio is very low. Furthermore, noise introduced by light source drift and optical components also affects the system's sensing performance. To improve the gain of the probe signal while reducing the impact of optical noise, this project proposes to use heterodyne detection technology to achieve phase acquisition of the OTDR signal. The light source is first split into probe light and local oscillator light by a coupler. The probe light is modulated into an optical pulse by an acousto-optic modulator (AOM) and an erbium-doped fiber amplifier (EDFA), and then injected into the sensing fiber after power amplification. The local oscillator light and the Rayleigh backscattered signal returning to the incident end form optical interference in another coupler, and are then converted by a balanced photodetector (BPD) and sampled by an ADC.

[0056] Considering the effects of optical frequency and frequency shift caused by AOM, this embodiment uses the following formula:

[0057] Rewritten as (ignoring polarization state):

[0058]

[0059] In the above formula, E R (t) represents the total Rayleigh backscattered signal after interference within the width of the optical pulse, A i (t), A(t), These represent the amplitude and phase of the Rayleigh backscattered signal generated at each scattering point, as well as its total amplitude and phase, respectively. f0 is the source frequency, f b The frequency shift introduced by the AOM (Aspect-Oriented Oscillator). Similarly, the local oscillator can be written as:

[0060]

[0061] E represents the amplitude of the local oscillator light. The initial phase of the local light. The splitting ratio of 2x2 coupler OC2 is usually 50:50, and such coupler has two characteristics, one is to ensure the total power of input and output optical signal unchanged, the output of each channel is the input in amplitude Two is to bring π / 2 frequency shift, the specific relationship is as follows:

[0062]

[0063] The output of BPD is two PDs respectively for two-way input photoelectric conversion and difference:

[0064]

[0065] Therefore, only the signal processing of BPD output electrical signal can realize the perception of fiber vibration state. In the construction of transformer fiber vibration system, the laser with small frequency drift and phase drift is used. Therefore, we usually think that the phase drift is approximately 0.

[0066] Looking at the relationship between phase change and fiber vibration, according to the relationship between optical phase difference and optical path difference, for two scattering points on the fiber, the phase difference can be expressed as:

[0067]

[0068] In the above formula, λ n is the wavelength of light propagating in the fiber with refractive index n, d ij represents the optical path difference of i, j two points on backscattering, z i , z j is the distance of two points from the starting end of the fiber, and the principle of two times of optical path difference is the same as that in OTDR.

[0069] Without considering the influence of temperature change on the optical parameters in the fiber, when the fiber is subjected to external strain, Poisson effect will cause the radial, tangential and axial expansion of the fiber, and photoelastic effect will cause the change of refractive index of the fiber. Ignoring the radial and tangential strain, formula:

[0070]

[0071] The differential can be obtained:

[0072]

[0073] Further analysis of axial expansion and refractive index change:

[0074]

[0075] Δ(z i -z j )=(zi - z j )ε

[0076] In the above two equations, p represents the axial photoelastic coefficient, and ε is the axial strain. It is obvious that there is a linear relationship between the phase and the strain:

[0077] Therefore, when the optical fiber is not affected by external strain, the phase difference between the two fixed points should remain unchanged. When subjected to stress, the phase difference between the two fixed points and the size of the strain are linearly related. There are complex stress fields, electromagnetic fields and temperature fields inside the transformer, which have an important influence on the image quality of the optical fiber sensing imaging, especially the positioning accuracy of the imaging points and the data quality. Poor positioning accuracy of the imaging points will cause changes in the texture and shape features of the image. The stability of the data quality of the imaging points directly determines the noise of the digital image. In order to ensure the reliability of the image features and improve the quality of the distributed optical fiber vibration imaging.

[0078] This embodiment uses heterodyne detection technology to obtain the phase of the OTDR signal, which improves the gain of the detection signal while reducing the influence of optical noise. The light source is first divided into detection light and local oscillator light by a coupler, where the detection light is modulated into optical pulses by an acousto-optic modulator AOM and an erbium-doped fiber amplifier EDFA and then power amplified before being injected into the sensing optical fiber. The local oscillator light and the Rayleigh backscattering signal returning to the incident end form optical interference in another coupler, and are converted by a balanced photodetector BPD and then sampled by an ADC. The control signal of the acousto-optic modulator is a pulse sequence given by a program-controlled function generator AWG. In addition, in terms of distributed optical fiber winding installation, a design form of independent winding of longitude and latitude is adopted to verify the spatial coordinates of the imaging points in the radial and axial directions, thereby improving the positioning accuracy of the imaging points.

[0079] After knowing the specific position of the vibration imaging point in the optical fiber, the time-domain backscattering signal at this position needs to be phase demodulated. This project plans to use the Hilbert phase demodulation method based on heterodyne interference to extract the phase difference and calculate the vibration waveform of the imaging point. The phase demodulation principle is shown in Figure 2 .

[0080] First, introduce two local carrier signals with a phase difference of 90° and a frequency shift f b as reference signals. Multiply one of the signals with the current signal in equation: to get:

[0081] At the same time, the I out (t) signal of equation is directly used as Q(t), and I out(t) is implemented by Hilbert transform, i.e.:

[0082]

[0083] where H[I out (t)] represents the Hilbert transform of I out (t). It can be seen that this method is more convenient than I / Q demodulation method.

[0084] S3, preprocessing the optical fiber vibration original image; the image preprocessing includes: image enhancement processing, image smoothing processing and image sharpening processing. The optical fiber vibration image reference Figure 3 is obtained after preprocessing.

[0085] The main purpose of feature extraction is to reduce the redundant information in the optical fiber vibration image, reduce the image dimension, and extract effective feature information to save time for the subsequent recognition stage. Various feature analysis methods have their own applicability, and the extraction of appropriate features is crucial for image recognition.

[0086] After obtaining the vibration information of the pixel points, 3Dmax modeling is used to import OpenGL to realize the visualization of the point cloud. Each pixel point is assigned to obtain the split optical fiber vibration original image containing the vibration information of the body.

[0087] In the process of optical fiber vibration imaging, there are image noise points caused by optical noise, phase ambiguity and other factors, therefore, the original image needs to be preprocessed. Specifically, the imaging image of the present subject mainly includes three basic links: grayscale processing, image enhancement and image segmentation.

[0088] Unlike the analysis of life scene images, the image database of industrial scenes is very limited, making it difficult to use deep learning methods based on convolutional neural networks for recognition and analysis. The embodiment extracts the image features x(t) = (C p ,E ρ ,I loc ,r c ,O c ), and constructs the insulation evaluation index H w (t).

[0089] S4, extracting the typical features of the preprocessed optical fiber vibration original image, specifically including extracting the typical features of the preprocessed optical fiber vibration original image based on the gray level co-occurrence matrix method, the typical features including texture contrast, optical fiber vibration image energy density, local texture variable, texture complexity and atlas correlation parameter.

[0090] The construction of the feature variable x. At present, the definable feature parameters are as follows:

[0091] Texture contrast:

[0092]

[0093] Optical fiber vibration image energy density:

[0094]

[0095] Local texture variable:

[0096]

[0097] Texture complexity:

[0098]

[0099] Atlas correlation parameter:

[0100]

[0101] Wherein, p is the gray level co-occurrence matrix element, n is the number of pixels, S i,j is the pixel area;

[0102]

[0103] S5, input the extracted typical features into the constructed transformer insulation evaluation model; the transformer insulation evaluation model is:

[0104]

[0105] Wherein, Y(x,t) is the state quantity of the monitoring image at time t, Y R (x,t0) is the state quantity at the initial time of insulation, x is the variable of typical features, Ω represents the imaging space of the optical fiber image, and adjusting the range thereof can detect the local or global of the transformer body.

[0106] S6, the transformer insulation evaluation model is solved to obtain the transformer insulation state data.

[0107] The key of the evaluation model construction is to find the optimal estimation of the unknown parameter vector , so that the evaluation model meets the physical mechanism of insulation aging shown on the image. After the state equation (evaluation model) is established and the characteristic parameters of the optical fiber vibration image are determined, the optimization of the evaluation model needs to determine the objective function and the optimization algorithm. Among them, the interference term of the parameter equation is the actual interference and the error of the evaluation model (the difference between the parameter vector and the optimal solution). As can be seen, the optimization object of the objective function should be the interference quantity in the evaluation model.

[0108] Suppose that in the optical fiber vibration image, the influence characteristic quantity x(t) = (C p , Eρ I loc ,r c ,O c ) of the interference model is ||w(x,t)||,||w B (ξ,t)||,||w φ (x)||, then the optimal estimate is to find the minimum value of the interference model and the error ||ε|| (ε = col(ε1,...,ε M )). At the same time, the objective function should have the corresponding appropriate weight for the norm of the random interference and the data vector extracted from the fiber image. Accordingly, the objective function is established as follows:

[0109]

[0110] where and represent the operators Ω dx i and is a symmetric positive definite weight coefficient matrix and has a generalized inverse, which can ensure that x(t) still has a smooth estimate in the case of low pixel density.

[0111] In this topic, the variational method is used as an optimization estimation algorithm to find the global minimum of the objective function. Because is the estimated value when the objective function takes the minimum value, the small perturbation increment δx should satisfy the following formula:

[0112]

[0113] Substitute equation (27) into equation (26), and combine with the state estimation model equation (17), we can get the equations and conditions that must be satisfied, see equations (28-29):

[0114]

[0115] where,

[0116]

[0117] M[δ] = col(M1[δ(u-u1)],...,M M [δ(u-u M )]) ; δu, δφ are small changes; C □ is the generalized inverse matrix of W □ ; K B (·) is an operator defined on the boundary .

[0118] To solve the boundary value problem, a suitable coordinate transformation is needed to convert the boundary value problem (BVP) into an initial value problem (IVP):

[0119] The transformation from BVP to IVP is accomplished by coordinate transformation equation (30), finding P(u,t) and The conversion can then be completed. The result of the solution is... Relevant equations Seeking Then, combined:

[0120]

[0121] The optimal estimate for the entire time domain is obtained by inverse time integration. That is, the evaluation model for insulation status is solved.

Claims

1. A method for online monitoring of transformer insulation condition based on fiber optic vibration imaging, characterized in that, Includes the following steps: S1, using distributed optical fiber to acquire vibration signals from the detection points of the transformer; S2, based on - The OTDR principle converts vibration signals into raw images of fiber optic vibration; S3, preprocess the original image of fiber vibration; S4, Extract typical features from the preprocessed original image of fiber vibration; S5. Input the extracted typical features into the constructed transformer insulation evaluation model; S6, The transformer insulation assessment model solves for typical characteristics to obtain transformer insulation status data; In step S2, the vibration signal is converted into the original image of optical fiber vibration using the Hilbert phase demodulation method based on heterodyne interference; the distributed optical fiber is wrapped around the transformer body using a latitude and longitude independent winding method; and the transformer insulation evaluation model is constructed based on the variational optimization algorithm.

2. The online monitoring method for transformer insulation status based on fiber optic vibration imaging according to claim 1, characterized in that, The S1 specifically includes the following steps: using distributed optical fibers to wrap around the transformer body to directly collect vibration signals from the transformer core and windings.

3. The online monitoring method for transformer insulation status based on fiber optic vibration imaging according to claim 1, characterized in that, S2 specifically includes the following steps: based on - OTDR's Rayleigh backscatter amplitude detection technology extracts the vibration signal at any point in global coordinates, and plots the original fiber vibration image based on the corresponding vibration signal in global coordinates. - The Rayleigh backscatter amplitude detection technology of the OTDR uses the Hilbert phase demodulation method with heterodyne interferometry.

4. The online monitoring method for transformer insulation status based on fiber optic vibration imaging according to claim 1, characterized in that, The preprocessing in S3 includes: grayscale conversion, image enhancement, and image segmentation.

5. The online monitoring method for transformer insulation status based on fiber optic vibration imaging according to claim 1, characterized in that, S4 specifically includes the following steps: extracting typical features of the preprocessed original optical fiber vibration image based on the gray-level co-occurrence matrix method. The typical features include texture contrast, optical fiber vibration image energy density, local texture variables, texture complexity, and spectral correlation parameters.

6. The online monitoring method for transformer insulation status based on fiber optic vibration imaging according to claim 1, characterized in that, The transformer insulation evaluation model is as follows: Where Y(x,t) is the state variable of the monitored image at time t, Y R (x,t0) represents the state variables at the initial moment of insulation, x is a variable of typical characteristics, and Ω represents the imaging space of the fiber optic image. Adjusting its range can be used to detect the local or global aspects of the transformer body. The transformer insulation evaluation model is optimized using the variational method.

7. A method for online monitoring of transformer insulation status based on fiber optic vibration imaging as described in claim 6, characterized in that, The variable x of the typical characteristics includes: Texture contrast: Energy density of fiber optic vibration images: Local texture variables: Texture complexity: Spectral correlation parameters: Where p is an element of the gray-level co-occurrence matrix, n is the number of pixels, and S i,j The pixel area;

Images