Oil paper insulation electrical aging stage identification method

Abstract

The invention relates to an oil paper insulation electrical aging stage identification method, which comprises: preparing an oil paper sample, collecting three-dimensional fluorescence spectrum data, carrying out an electrical aging experiment on the oil paper sample, collecting the oil paper sample after partial discharge, and determining the three-dimensional fluorescence spectrum data; performing parallel factor analysis on the three-dimensional fluorescence spectrum data, extracting characteristic parameters of fluorescence components, establishing a parallel factor analysis model based on the three-dimensional fluorescence spectrum data, and performing characteristic parameter identification and extraction on the fluorescence components in the parallel factor analysis model; an electrical aging stage identification model based on the characteristic parameters is established, clustering analysis is performed on the characteristic parameters based on a K clustering algorithm, and the electrical aging stage identification model is established; the characteristic parameters are input into an electrical aging stage recognition model, the electrical aging stage of the oil paper insulation is output, new characteristic parameters are input into the electrical aging stage recognition model, and the judgment result of the electrical aging stage of the oil paper insulation is output.

Description

Technical Field

[0001] The present invention relates to the technical field of insulation state recognition of power equipment, and particularly to a method for recognizing the electrical aging stage of oil-paper insulation. Background Art

[0002] As a key insulation structure of high-voltage electrical equipment such as power transformers, the performance of oil-paper insulation is directly related to the operation safety and reliability of the equipment. During long-term operation, oil-paper insulation will be gradually aged under the combined action of various factors such as electric field, temperature, and humidity, resulting in a decline in insulation performance, which may cause equipment failures and seriously affect the safe and stable operation of the power system. At present, the commonly used methods for evaluating the aging state of oil-paper insulation include dissolved gas analysis in oil, dielectric spectroscopy measurement, etc. However, these methods still have certain limitations in terms of the recognition accuracy and convenience in the electrical aging stage. Although dissolved gas analysis in oil can reflect abnormal conditions such as thermal decomposition and arc discharge generated during the aging process of insulation materials, its results are easily affected by factors such as the temperature, pressure, and solubility of the oil, and it is not sensitive enough for the recognition of early aging states. Dielectric spectroscopy measurement can provide relatively comprehensive insulation information, but it requires professional measurement equipment and complex testing processes, which are restricted in on-site applications, and it is difficult to accurately identify complex aging states with a single dielectric spectroscopy parameter.

[0003] Three-dimensional fluorescence spectroscopy has high sensitivity and resolution, can detect changes in trace amounts of fluorescent substances, has obvious advantages for the recognition of early aging states, and has the advantages of non-destructiveness and rapid detection, and is suitable for on-site rapid assessment of the aging state of oil-paper insulation. However, how to effectively extract the characteristic parameters from the three-dimensional fluorescence spectroscopy and accurately correspond them to the electrical aging stage of oil-paper insulation is still an urgent problem to be solved.

[0004] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present invention, and therefore may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0005] The present invention provides a method for identifying the electrical aging stage of oil-paper insulation. By collecting the three-dimensional fluorescence spectrum data of an oil-paper insulation sample, applying parallel factor analysis to extract feature vectors, and combining with the K-clustering algorithm to establish an identification model, accurate identification of the electrical aging stage of oil-paper insulation is achieved. Parallel factor analysis (PARAFAC), as a multi-dimensional data decomposition method, can decompose complex three-dimensional fluorescence spectrum data into the contributions of several fluorescence components, obtaining the excitation wavelength, emission wavelength, and corresponding intensity information of each fluorescence component. These information constitute the characteristic parameters of the fluorescence components. Parallel factor analysis can effectively handle the overlapping peaks and noise problems in three-dimensional fluorescence spectrum data, improving the accuracy and reliability of the data, and this method establishes the corresponding relationship between accurate characteristic parameters and the aging stage. This method has high sensitivity and high accuracy, providing strong support for the maintenance and management of power equipment.

[0006] A method for identifying the electrical aging stage of oil-paper insulation includes:

[0007] Step S1: Prepare an oil-paper sample and collect three-dimensional fluorescence spectrum data. Among them, an electrical aging experiment is carried out on the oil-paper sample, and the oil-paper sample after partial discharge is collected for the determination of three-dimensional fluorescence spectrum data;

[0008] Step S2: Perform parallel factor analysis on the three-dimensional fluorescence spectrum data to extract the characteristic parameters of the fluorescence components. Among them, a parallel factor analysis model is established based on the three-dimensional fluorescence spectrum data, and the characteristic parameter identification and extraction of the fluorescence components in the parallel factor analysis model are carried out;

[0009] Step S3: Establish an electrical aging stage identification model based on the characteristic parameters. Among them, cluster analysis is carried out on the characteristic parameters based on the K-clustering algorithm to build an electrical aging stage identification model;

[0010] Step S4: Input the characteristic parameters into the electrical aging stage identification model and output the electrical aging stage of the oil-paper insulation. Among them, new characteristic parameters are input into the electrical aging stage identification model, and the determination result of the electrical aging stage of the oil-paper insulation is output.

[0011] In the described method for identifying the electrical aging stage of oil-paper insulation, in step S1, the form of partial discharge is needle-plate discharge, and the distance between the tip of the needle electrode and the plate electrode is 2 mm.

[0012] In the described method for identifying the electrical aging stage of oil-paper insulation, an oil-paper sample with a diameter of 80 mm × a thickness of 1 mm is placed between the needle electrode and the plate electrode. The material of the plate electrode is brass, with a diameter of 75 mm and a thickness of 15 mm; the material of the needle electrode is pure tungsten, and its curvature radius is 50 μm.

[0013] In the described method for identifying the electrical aging stage of oil-paper insulation, in step S1, the step-up method is used to determine the partial discharge inception voltage PDIV. During the electrical aging process, the positive and negative voltages are maintained at 1.5 times the inception voltage PDIV. Insulating oil is taken once per hour until the insulating paper is broken down. During the electrical aging process, a three-way valve and an oil pump are used to take oil without power interruption.

[0014] In the described method for identifying the electrical aging stage of oil-paper insulation, in step S1, when collecting three-dimensional fluorescence spectrum data, the slit 1 of the steady-state and transient fluorescence spectrometer is set to 5 nm, and the slit 2 is set to 0.3 nm; the excitation wavelength range is set to 330 - 450 nm, and the emission wavelength range is set to 350 - 750 nm; the excitation wavelength scanning interval is set to 5 nm, and the emission wavelength scanning interval is set to 2 nm; the spectrometer scanning dwell time is set to 0.02 s.

[0015] In the described method for identifying the electrical aging stage of oil-paper insulation, the characteristic parameters of the fluorescent components include excitation load, emission load, and relative fluorescence intensity.

[0016] In the described method for identifying the electrical aging stage of oil-paper insulation, step S2 includes

[0017] Step S21: Preprocess the collected three-dimensional fluorescence spectrum data. The preprocessing includes baseline correction, smoothing, and normalization;

[0018] Step S22: Select the number of components R by comparing the core consistency, residual value, and model interpretation rate of the parallel factor analysis model;

[0019] Step S23: Iterative decomposition: Initialize the eigenvectors: the excitation wavelength eigenvector a k , the emission wavelength eigenvector b k and the sample pattern eigenvector c k , use the least squares error of the parallel factor analysis model as the objective function, and use the alternating least squares method ALS for iterative decomposition. By alternately optimizing the eigenvectors, gradually approach the minimum value of the objective function. When the objective function reaches a stable state, the iteration stops;

[0020] Step S24: After decomposition, obtain R groups of excitation wavelength eigenvectors a k , emission wavelength eigenvectors b k and sample pattern eigenvectors c k .

[0021] In the described method for identifying the electrical aging stage of oil-paper insulation, step S3 includes:

[0022] Step S31: Initialization of the clustering center: Select the comprehensive eigenvectors (a k , bk , c k ), as the initial clustering centers, so that the clustering centers represent different characteristic distributions of the data;

[0023] Step S32: Sample distance calculation: In the process of iteratively assigning samples, use the Euclidean distance formula to calculate the distance between the sample and the clustering center. For sample i and clustering center j, their comprehensive feature vectors are (a i , b i , c i ) and (a j , b j , c j ), and the corresponding Euclidean distance formula is:

[0024] ,

[0025] where, represents the distance of the excitation wavelength feature vector, represents the distance of the emission wavelength feature vector, represents the distance of the sample pattern feature vector;

[0026] Step S33: Clustering center update: Recalculate the new clustering center according to the feature vectors of all samples in the same cluster. For cluster j, the new clustering center is expressed as:

[0027] ,

[0028] where, is the sample set in cluster j, a i , b i , c i are the characteristic parameters of sample i,

[0029] Step S34: Evaluation and analysis of the electrical aging stage recognition model: By comparing the feature vectors in different clusters, reveal the characteristic differences of oil-paper insulation in different electrical aging stages.

[0030] In the described method for recognizing the electrical aging stage of oil-paper insulation, in step S4, the comprehensive evaluation parameter E is composed of the feature vectors a k , b k , c k , and the calculation formula is:

[0031] ,

[0032] where: a k , b k and c k respectively represent the excitation wavelength, emission wavelength and sample pattern feature vectors; a min , b minand c min respectively represent the minimum values of a k , b k and c k ; a max , b max and c max respectively represent the maximum values of a k , b k and c k ; , and are weight coefficients. By calculating the comprehensive evaluation parameter E, the determination result of the electrical aging stage of oil-paper insulation is output.

[0033] In the method for identifying the electrical aging stage of oil-paper insulation described above, when 0.1 ≤ E < 0.3, the electrical aging stage of oil-paper insulation is output as the early aging stage; when 0.3 ≤ E < 0.6, the electrical aging stage of oil-paper insulation is output as the middle aging stage; when 0.6 ≤ E < 0.9, the electrical aging stage of oil-paper insulation is output as the late aging stage.

[0034] Compared with the prior art, the present invention has the following advantages:

[0035] (1) The present invention combines three-dimensional fluorescence spectroscopy with parallel factor analysis, which can effectively extract the trace fluorescence characteristic information in the aging process of oil-paper insulation. The three-dimensional fluorescence spectroscopy technology has high sensitivity and can detect the trace fluorescent substances generated during the electrical aging of oil-paper insulation, which are often difficult to be found in traditional detection methods. Parallel factor analysis further decomposes the complex three-dimensional fluorescence spectroscopy data into the characteristic vectors of multiple fluorescence components, and extracts the key characteristic information of the excitation wavelength, emission wavelength and sample mode, thereby improving the accuracy of identification;

[0036] (2) The three-dimensional fluorescence spectroscopy adopted by the present invention is a non-destructive detection method, which will not cause any damage to the oil-paper insulation sample and is suitable for on-site rapid detection. The three-dimensional fluorescence spectrometer can quickly obtain the fluorescence spectroscopy data of the oil sample. Combining parallel factor analysis and K-clustering algorithm, it can complete the identification of the electrical aging stage in a short time. This feature makes the present invention have significant advantages in practical applications. It can quickly obtain the aging state information of oil-paper insulation without affecting the normal operation of the equipment, providing timely decision-making support for the maintenance and management of power equipment;

[0037] (3) By extracting the excitation wavelength, emission wavelength, and sample mode feature vectors, the present invention comprehensively evaluates the electrical aging state of oil-paper insulation from multiple dimensions. The definition and calculation method of the comprehensive evaluation parameter E make the evaluation of the electrical aging state more scientific and quantitative. The present invention has established a clear correspondence between the parameter values and the electrical aging stages, providing a reliable scientific basis for the condition assessment of power equipment. This comprehensive evaluation method can not only accurately identify the aging stages of oil-paper insulation but also provide detailed reference data for the maintenance and management of power equipment, helping to formulate more reasonable maintenance strategies.

[0038] (4) The present invention can accurately identify the electrical aging stages of oil-paper insulation, timely detect potential insulation faults, and take corresponding preventive measures, thereby reducing maintenance costs. By identifying the oil-paper insulation in the early stage of aging in advance, appropriate maintenance measures can be taken to extend the service life of the equipment and reduce the economic losses caused by equipment failures. At the same time, the rapid detection feature of the present invention can ensure the safe and stable operation of power equipment and improve the operation efficiency of the power system. Accurately identifying the aging state of oil-paper insulation avoids unnecessary equipment shutdowns and repairs, ensures the continuous power supply of the power system, and provides strong support for the development of the social economy. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] By reading the following detailed description of the preferred embodiments, various other advantages and benefits of the present invention will become clear to those of ordinary skill in the art. The accompanying drawings in the specification are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. Moreover, throughout the drawings, the same reference numerals are used to represent the same components.

[0040] In the drawings:

[0041] Figure 1 is a schematic flow chart of a method for identifying the electrical aging stage of oil-paper insulation based on the parallel factor characteristic parameters of three-dimensional fluorescence spectroscopy according to the present invention;

[0042] Figure 2 is a contour map drawn from the three-dimensional fluorescence spectroscopy data of FR3 vegetable insulating oil collected in a specific embodiment of a method for identifying the electrical aging stage of oil-paper insulation based on the parallel factor characteristic parameters of three-dimensional fluorescence spectroscopy according to the present invention;

[0043] Figure 3 is a graph showing the variation of the relative fluorescence intensity characteristic parameters of three fluorescence components of FR3 vegetable insulating oil with the electrical aging time in a specific embodiment of a method for identifying the electrical aging stage of oil-paper insulation based on the parallel factor characteristic parameters of three-dimensional fluorescence spectroscopy according to the present invention;

[0044] Figure 4 It is a graph showing the corresponding relationship between the comprehensive evaluation parameter E, which is a specific embodiment of the method for identifying the electrical aging stage of oil-paper insulation based on the parallel factor characteristic parameters of three-dimensional fluorescence spectroscopy in the present invention, and the electrical aging stage of oil-paper insulation.

[0045] The following further explains the present invention in conjunction with the drawings and embodiments. Specific Embodiment

[0046] The specific embodiments of the present invention will be described in more detail below with reference to the drawings. Although the specific embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0047] It should be noted that certain terms are used in the specification and claims to refer to specific components. Those skilled in the art should understand that technicians may use different terms to refer to the same component. The specification and claims do not use the difference in terms as a way to distinguish components, but use the difference in the functions of components as the criterion for distinction. For example, the terms "comprising" or "including" mentioned throughout the specification and claims are open-ended terms and should be interpreted as "including but not limited to". The subsequent description in the specification is the preferred embodiment for implementing the present invention, but the description is for the purpose of the general principles of the specification and is not intended to limit the scope of the present invention. The protection scope of the present invention shall be defined by the appended claims.

[0048] For the convenience of understanding the embodiments of the present invention, the following will further explain with specific embodiments as examples in conjunction with the drawings, and each drawing does not constitute a limitation to the embodiments of the present invention.

[0049] As Figures 1 to 4 shown, the method for identifying the electrical aging stage of oil-paper insulation includes the following steps:

[0050] Step S1: Prepare an oil-paper sample and collect three-dimensional fluorescence spectroscopy data. Among them, an electrical aging experiment is carried out on the oil-paper sample, and the three-dimensional fluorescence spectroscopy data of the oil-paper sample after partial discharge is collected for measurement.

[0051] Step S2: Perform parallel factor analysis on the three-dimensional fluorescence spectroscopy data to extract the characteristic parameters of the fluorescent components. Among them, a parallel factor analysis model is established based on the three-dimensional fluorescence spectroscopy data, and the characteristic parameter identification and extraction of the fluorescent components in the parallel factor analysis model are carried out.

[0052] Step S3: Establish an identification model for the electrical aging stage based on characteristic parameters. Among them, perform clustering analysis on the characteristic parameters based on the K-clustering algorithm to build an identification model for the electrical aging stage;

[0053] Step S4: Input the characteristic parameters into the identification model for the electrical aging stage, and output the electrical aging stage of the oil-paper insulation. Among them, input new characteristic parameters into the identification model for the electrical aging stage, and output the determination result of the electrical aging stage of the oil-paper insulation.

[0054] In the preferred implementation manner of the method for identifying the electrical aging stage of oil-paper insulation, in step S1, the form of partial discharge is needle-plate discharge, and the distance between the tip of the needle electrode and the plate electrode is 2 mm.

[0055] In the preferred implementation manner of the method for identifying the electrical aging stage of oil-paper insulation, an oil-paper sample with a diameter of 80 mm × a thickness of 1 mm is placed between the needle electrode and the plate electrode. The material of the plate electrode is brass, with a diameter of 75 mm and a thickness of 15 mm; the material of the needle electrode is pure tungsten, and its curvature radius is 50 μm.

[0056] In the preferred implementation manner of the method for identifying the electrical aging stage of oil-paper insulation, in step S1, the step-up voltage method is used to determine the partial discharge inception voltage PDIV. During the electrical aging process, the positive and negative voltages are maintained at 1.5 times the inception voltage PDIV. Insulating oil is taken once per hour until the insulating paper is broken down. During the electrical aging process, a three-way valve and an oil pump are used to take oil without power interruption.

[0057] In the preferred implementation manner of the method for identifying the electrical aging stage of oil-paper insulation, in step S1, when collecting three-dimensional fluorescence spectrum data, the slit 1 of the steady-state and transient fluorescence spectrometer is set to 5 nm, and the slit 2 is set to 0.3 nm; the excitation band range is set to 330 - 450 nm, and the emission band range is set to 350 - 750 nm; the excitation band scanning interval is set to 5 nm, and the emission band scanning interval is set to 2 nm; the scanning residence time of the spectrometer is set to 0.02 s.

[0058] In the preferred implementation manner of the method for identifying the electrical aging stage of oil-paper insulation, the characteristic parameters of the fluorescent components include excitation load, emission load, and relative fluorescence intensity.

[0059] In the preferred implementation manner of the method for identifying the electrical aging stage of oil-paper insulation, step S2 includes

[0060] Step S21: Preprocess the collected three-dimensional fluorescence spectrum data. The preprocessing includes baseline correction, smoothing processing, and normalization;

[0061] Step S22: Select the number of components R by comparing the core consistency, residual value, and model interpretation rate of the parallel factor analysis model;

[0062] Step S23: Iterative decomposition: Initialize the feature vectors a k , b k , c k , set the objective function as the least squares error of the model, and use the alternating least squares method ALS for iterative decomposition. When the objective function reaches a steady state, the iteration stops;

[0063] Step S24: After decomposition, obtain R groups of feature vectors a k , b k , c k , which respectively correspond to the characteristics of the excitation load, emission load, and component relative fluorescence intensity.

[0064] In the preferred implementation of the above-mentioned method for identifying the electrical aging stage of oil-paper insulation, the step S3 includes:

[0065] Step S31: Initialization of cluster centers: Select the comprehensive feature vectors (a k , b k , c k ) of different components k as the initial cluster centers, so that the cluster centers represent different characteristic distributions of the data;

[0066] Step S32: Calculation of sample distance: In the process of iteratively assigning samples, use the Euclidean distance formula to calculate the distance between the sample and the cluster center. For sample i and cluster center j, their comprehensive feature vectors are respectively (a i , b i , c i ) and (a j , b j , c j ), and the corresponding Euclidean distance formula is:

[0067] ,

[0068] where represents the distance of the excitation wavelength feature vector, represents the distance of the emission wavelength feature vector, represents the distance of the sample pattern feature vector;

[0069] Step S33: Update of cluster centers: Recalculate the new cluster centers according to the feature vectors of all samples within the same cluster. For cluster j, the new cluster center is expressed as:

[0070] ,

[0071] where is the set of samples in cluster j, a i , bi ,c i is the characteristic parameter of sample i

[0072] Step S34: Evaluation and analysis of the electrical aging stage recognition model: By comparing the feature vectors in different clusters, the characteristic differences of oil-paper insulation at different electrical aging stages are revealed.

[0073] In the preferred embodiment of the method for identifying the electrical aging stage of oil-paper insulation, in step S4, the comprehensive evaluation parameter E is composed of the feature vectors a k ,b k ,c k and the calculation formula is:

[0074] ,

[0075] where: a k ,b k and c k represent the excitation wavelength, emission wavelength and sample mode feature vector respectively; a min ,b min and c min represent the minimum values of a k ,b k and c k respectively; a max ,b max and c max represent the maximum values of a k ,b k and c k respectively; 、 and are the weight coefficients. By calculating the comprehensive evaluation parameter E, the determination result of the electrical aging stage of oil-paper insulation is output.

[0076] In the preferred embodiment of the method for identifying the electrical aging stage of oil-paper insulation, when 0.1 ≤ E < 0.3, the electrical aging stage of oil-paper insulation is output as the early aging stage; when 0.3 ≤ E < 0.6, the electrical aging stage of oil-paper insulation is output as the middle aging stage; when 0.6 ≤ E < 0.9, the electrical aging stage of oil-paper insulation is output as the late aging stage.

[0077] In one embodiment

[0078] Example 1

[0079] A method for identifying the electrical aging stage of oil-paper insulation includes:

[0080] S1 Prepare an insulating oil sample and collect three-dimensional fluorescence spectral data: Build an oil-paper insulation model using pre-treated insulating oil and insulating cardboard and conduct an electrical aging experiment. Collect the insulating oil sample after partial discharge for the determination of three-dimensional fluorescence spectral data. In step S1, the form of partial discharge is needle-plate discharge. The plate electrode material is brass, with a diameter of 75 mm and a thickness of 15 mm. The needle electrode material is pure tungsten, with a curvature radius of 50 μm. In the oil-paper insulation defect model, a circular insulating cardboard sample with a diameter of 80 mm (diameter) × 1 mm (thickness) is placed between the high-voltage electrode and the ground electrode. The distance from the tip of the needle electrode to the lower plate electrode or insulating cardboard is 2 mm. During the experiment, the stepped voltage increase method is used. During the pre-experiment, determine the partial discharge inception voltage PDIV of the oil-paper insulation system. During the electrical aging process, the positive and negative voltages are maintained at 1.5 times PDIV. Take insulating oil once per hour until the insulating cardboard is broken down. During the electrical aging process, a three-way valve and an oil pump are used to take oil without power interruption, avoiding the influence of power interruption on the electrical aging experiment. The parameters set during the collection of three-dimensional fluorescence spectral data are as follows: The slit 1 of the steady-state transient fluorescence spectrometer is set to 5 nm, and the slit 2 is set to 0.3 nm. The excitation band range is set to 330 - 450 nm, and the emission band range is set to 350 - 750 nm. The excitation band scanning interval is set to 5 nm, and the emission band scanning interval is set to 2 nm. The spectrometer scanning dwell time is set to 0.02 s.

[0081] S2 Perform parallel factor analysis on the three-dimensional fluorescence spectral data to extract the characteristic parameters of the fluorescent components: Use the pre-treated three-dimensional fluorescence spectral data to establish a parallel factor analysis model, and identify and extract the characteristic parameters of the fluorescent component information in the model. The characteristic parameters of the fluorescent components in step S2 include excitation loading, emission loading, and relative fluorescence intensity. The specific steps of step S2 are as follows:

[0082] S21: Pre-treat the collected three-dimensional fluorescence spectral data, including baseline correction, smoothing, and normalization, etc., to ensure the accuracy and comparability of the data;

[0083] S22: Select the appropriate number of components R by comparing the core consistency, residual value, and model interpretation rate of the parallel factor analysis model;

[0084] S23: Iterative decomposition: Initialize the eigenvectors a k , b k , c k , set the objective function as the least square error of the model, and use the alternating least squares method (ALS) for iterative decomposition. When the objective function reaches a stable state, the iteration stops;

[0085] S24: After decomposition, obtain R groups of eigenvectors a k , b k , ck , corresponding to the characteristics of the excitation load, emission load, and relative fluorescence intensity of the components respectively.

[0086] S3 Establish an identification model for the electrical aging stage based on characteristic parameters: Based on the K-clustering algorithm, perform clustering analysis on the fluorescence component characteristic parameters obtained from the parallel factor analysis model, and build an identification model for the electrical aging stage of oil-paper insulation; the K-clustering algorithm is used to establish the electrical aging stage identification model in step S3; the specific steps of step S3 are as follows:

[0087] S31: Initialization of clustering centers: Select the comprehensive feature vectors (a k , b k , c k ) of different components k as the initial clustering centers, so that the clustering centers can represent different characteristic distributions of the data, thereby improving the stability and accuracy of clustering;

[0088] S32: Calculation of sample distances: In the process of iteratively assigning samples, use the Euclidean distance formula to calculate the distance between the sample and the clustering center. For sample i and clustering center j, their comprehensive feature vectors are respectively (a i , b i , c i ) and (a j , b j , c j ), and the corresponding Euclidean distance formula is:

[0089]

[0090] where, represents the distance of the excitation wavelength feature vector, represents the distance of the emission wavelength feature vector, represents the distance of the sample pattern feature vector.

[0091] S33: Update of clustering centers: Recalculate the new clustering centers according to the feature vectors of all samples within the same cluster. For cluster j, the new clustering center can be expressed as:

[0092]

[0093] where, is the sample set in cluster j, a i , b i , c i are the characteristic parameters of sample i.

[0094] S34: Model evaluation and analysis: By comparing the feature vectors in different clusters, reveal the characteristic differences of oil-paper insulation in different electrical aging stages.

[0095] S4 inputs the characteristic parameters into the recognition model and outputs the electrical aging stage of the oil-paper insulation: inputs the new fluorescence component characteristic parameters into the oil-paper insulation electrical aging stage recognition model obtained in step S3, calculates the comprehensive evaluation parameter E, and outputs the determination result of the electrical aging stage of the oil-paper insulation; the comprehensive evaluation parameter E in step S4 consists of the feature vectors a k , b k , c k , and the calculation formula is:

[0096]

[0097] where: a k , b k , and c k represent the excitation wavelength, emission wavelength, and sample mode feature vector respectively; a min , b min , and c min represent the minimum values of a k , b k , and c k respectively; a max , b max , and c max represent the maximum values of a k , b k , and c k respectively; , , and are weight coefficients, which can be adjusted according to experimental data and actual requirements.

[0098] According to a large amount of experimental data and statistical analysis, the following corresponding relationships can be formulated:

[0099] Specific embodiments

[0100] Using Widman insulating cardboard with a thickness of 1 mm and FR3 plant insulating oil, the steps of the oil-paper insulation electrical aging stage recognition method are as follows:

[0101] S1 prepares FR3 plant insulating oil samples and collects three-dimensional fluorescence spectrum data: uses the pre-treated FR3 plant insulating oil and Widman insulating cardboard with a thickness of 1 mm to build an oil-paper insulation model and conducts an electrical aging experiment, collects the insulating oil samples after partial discharge for the determination of three-dimensional fluorescence spectrum data, and uses the contour map drawn by the three-dimensional fluorescence spectrum data of FR3 plant insulating oil, as Figure 2 shown;

[0102] S2 performs parallel factor analysis on the three-dimensional fluorescence spectral data of FR3 plant insulating oil to extract the characteristic parameters of the fluorescent components: uses the preprocessed three-dimensional fluorescence spectral data of FR3 plant insulating oil to establish a parallel factor analysis model, and identifies and extracts the characteristic parameters of the fluorescent component information in the model; the characteristic parameters of the fluorescent components in step S2 include excitation loading, emission loading, and relative fluorescence intensity; the specific steps of step S2 are as follows:

[0103] Step S2 includes

[0104] Step S21: Preprocess the collected three-dimensional fluorescence spectral data, and the preprocessing includes baseline correction, smoothing, and normalization;

[0105] Step S22: Select the number of components R by comparing the core consistency, residual value, and model interpretation rate of the parallel factor analysis model;

[0106] Step S23: Iterative decomposition: Initialize the eigenvectors: excitation wavelength eigenvector a k emission wavelength eigenvector b k and sample pattern eigenvector c k , use the least squares error of the parallel factor analysis model as the objective function, and perform iterative decomposition using the alternating least squares method ALS. By alternately optimizing the eigenvectors, gradually approach the minimum value of the objective function. When the objective function reaches a stable state, the iteration stops;

[0107] Step S24: After decomposition, obtain R groups of excitation wavelength eigenvectors a k emission wavelength eigenvectors b k and sample pattern eigenvectors c k , which respectively correspond to the characteristics of excitation loading, emission loading, and component relative fluorescence intensity. The variation diagram of the relative fluorescence intensity characteristic parameters of the three fluorescent components of FR3 plant insulating oil with the electrical aging time is as Figure 3 shown.

[0108] S3 establishes an electrical aging stage identification model based on three groups of characteristic parameters: Based on the K-clustering algorithm, perform clustering analysis on the characteristic parameters of the fluorescent components obtained from the parallel factor analysis model, and build an electrical aging stage identification model for oil-paper insulation; in step S3, the K-clustering algorithm is used to establish an electrical aging stage identification model; the specific steps of step S3 are as follows:

[0109] S31: Initialize the clustering center: Select the comprehensive eigenvector (a k , b k , c k ) of different components k as the initial clustering center, so that the clustering center can represent different characteristic distributions of the data, thereby improving the stability and accuracy of clustering;

[0110] S32: Sample distance calculation: During the process of iteratively allocating samples, the Euclidean distance formula is used to calculate the distance between the sample and the cluster center. For sample i and cluster center j, their comprehensive feature vectors are (a i , b i , c i ) and (a j , b j , c j ) respectively, and the corresponding Euclidean distance formula is:

[0111]

[0112] where, represents the distance of the excitation wavelength feature vector, represents the distance of the emission wavelength feature vector, represents the distance of the sample pattern feature vector.

[0113] S33: Cluster center update: The new cluster center is recalculated according to the feature vectors of all samples within the same cluster. For cluster j, the new cluster center can be expressed as:

[0114]

[0115] where, is the sample set in cluster j, and a i , b i , c i are the characteristic parameters of sample i.

[0116] S34: Model evaluation and analysis: By comparing the feature vectors in different clusters, the characteristic differences of oil-paper insulation at different electrical aging stages are revealed.

[0117] S4 Input the characteristic parameters into the recognition model and output the electrical aging stage of the oil-paper insulation: Input the new fluorescence component characteristic parameters into the oil-paper insulation electrical aging stage recognition model obtained in step S3, and by calculating the comprehensive evaluation parameter E, output the determination result of the electrical aging stage of the oil-paper insulation; the comprehensive evaluation parameter E in step S4 is composed of the feature vectors a k , b k , c k , and the calculation formula is:

[0118]

[0119] where: a k , b k and c k represent the excitation wavelength, emission wavelength, and sample pattern feature vectors respectively; a min , b min and cmin respectively represent the minimum values of a k , b k and c k ; a max , b max and c max respectively represent the maximum values of a k , b k and c k ; , and are weight coefficients and can be adjusted according to experimental data and actual requirements.

[0120] Based on a large amount of experimental data and statistical analysis, the following corresponding relationships can be formulated:

[0121]

[0122] In this specific embodiment , , .

[0123] Experimental test results show that the comprehensive evaluation parameter E of the FR3 oil-paper insulation system approximately shows a linear growth trend with the increase of the electrical aging time. When the electrical aging time is 0 ≤ t < 2 h, that is, in the early stage of electrical aging, the comprehensive evaluation parameter E satisfies 0.1 ≤ E < 0.3; when the electrical aging time is 2 ≤ t < 4 h, that is, in the middle stage of electrical aging, the comprehensive evaluation parameter E satisfies 0.3 ≤ E < 0.6; when the electrical aging time is 4 ≤ t < 6 h, that is, in the late stage of electrical aging, the comprehensive evaluation parameter E satisfies 0.6 ≤ E < 0.9. The corresponding relationship between the comprehensive evaluation parameter E and the electrical aging stage of the oil-paper insulation is as Figure 4 shown.

[0124] In this embodiment, through the electrical aging experiment on the FR3 oil-paper insulation system, the change of the comprehensive evaluation parameter E under different electrical aging times is tested. The experimental results show that the comprehensive evaluation parameter E approximately shows a linear growth trend with the increase of the electrical aging time, and can accurately reflect different electrical aging stages of the oil-paper insulation system, providing strong support for the condition assessment and maintenance of power equipment.

[0125] In one embodiment, in step S2,

[0126] (1) Data preprocessing:

[0127] Baseline correction: Eliminate the baseline drift in the fluorescence spectrum to ensure the accuracy of the spectral data;

[0128] Smoothing processing: Use a smoothing algorithm to reduce the random noise in the data and improve the signal-to-noise ratio of the data;

[0129] Normalization: Normalize the spectral data so that it is within the same scale range, facilitating subsequent analysis and comparison.

[0130] (2) Model parameter selection:

[0131] Core consistency comparison: Core consistency is used to measure the stability and reliability of the model. The closer its value is to 100%, the better. Generally, a model with a core consistency higher than 90% is considered reliable and can be used for further analysis. If the core consistency is lower than 90%, it indicates that the model may be unstable and the number of components or other model parameters need to be reconsidered. By comparing the core consistency of the model under different numbers of components, select the most appropriate core consistency value to ensure the stability and reliability of the model;

[0132] Residual value analysis: The residual value reflects the goodness of fit of the model to the data. The smaller the residual value, the better the fitting effect of the model. When selecting the number of components, a model with a smaller residual value should be chosen. If the residual value is too large, it means that the model fails to fully explain the variation in the data and the number of components may need to be increased. Evaluate the residual value of the model and select the number of components with a smaller residual value to improve the fitting degree of the model;

[0133] Model interpretability evaluation: The model interpretability represents the proportion of data variation that the model can explain. The closer its value is to 100%, the better. Generally, a model with an interpretability higher than 80% is considered good and can be used for practical analysis. If the model interpretability is lower than 80%, it indicates that the model may fail to fully capture the main features in the data and the number of components or other model parameters need to be reconsidered. Select the number of components that can explain most of the data variation to ensure that the model has good interpretability.

[0134] (3) Iterative decomposition:

[0135] Initialize the feature vectors: Initialize the feature vectors a k (excitation wavelength feature vector), b k (emission wavelength feature vector), and c k (sample pattern feature vector);

[0136] Set the objective function: Take the least squares error of the model as the objective function, aiming to minimize the difference between the model prediction value and the actual value;

[0137] Alternating Least Squares (ALS): Use the ALS algorithm for iterative decomposition. By alternately optimizing the feature vectors, gradually approach the minimum value of the objective function. When the objective function reaches a stable state, the iteration stops.

[0138] (4) Feature parameter extraction:

[0139] Excitation wavelength feature vector a k: Containing different intensity values within the excitation wavelength range, whose peak position corresponds to the main excitation wavelength of component k,

[0140] Emission wavelength eigenvector b k : Containing different intensity values within the emission wavelength range, whose peak position corresponds to the main emission wavelength of component k,

[0141] Sample pattern eigenvector c k : Containing the relative intensity of this component in each sample, which can reveal the change trend of this component in oil-paper insulation samples at different aging stages.

[0142] In one embodiment, S34: Model evaluation and analysis: By comparing the eigenvectors in different clusters, the characteristic differences of oil-paper insulation at different electrical aging stages are revealed, including,

[0143] (1) Eigenvector comparison:

[0144] Compare the eigenvectors in each cluster (excitation wavelength eigenvector a k , emission wavelength eigenvector b k and sample pattern eigenvector c k ) in detail, and analyze the differences of eigenvectors between different clusters. By comparing the eigenvectors of different clusters, specifically referring to the combination of specific excitation wavelength and emission wavelength in this patent, as well as the change trend of the sample pattern eigenvector at different aging stages, the typical characteristics of oil-paper insulation at different electrical aging stages can be identified;

[0145] (2) Characteristic difference revelation:

[0146] Combine the comparison results of eigenvectors to reveal the characteristic differences of oil-paper insulation at different electrical aging stages. For example, in the early stage of aging, the relative intensity of some fluorescent components is relatively low, while in the late stage of aging, the relative intensity of these components may increase significantly. By analyzing these characteristic differences, the corresponding relationship between the electrical aging stage and the characteristic parameters of fluorescent components can be established, providing a basis for the subsequent identification of electrical aging stages.

[0147] (3) Model optimization and verification:

[0148] According to the analysis results of characteristic differences, optimize and verify the electrical aging stage identification model to ensure that the model can accurately identify oil-paper insulation samples at different electrical aging stages. The accuracy and reliability of the model can be further improved by increasing training samples, adjusting clustering parameters, etc.

[0149] Although the embodiments of the present invention have been described above in conjunction with the accompanying drawings, the present invention is not limited to the above specific embodiments and application fields. The above specific embodiments are merely illustrative and guiding, rather than restrictive. Those of ordinary skill in the art can also make many forms under the inspiration of this specification and without departing from the scope protected by the claims of the present invention, and all of these fall within the scope of protection of the present invention.

Claims

1. A method for identifying the electrical aging stage of oil-paper insulation, characterized in that It includes the following steps: Step S1: Prepare an oil-paper sample and collect three-dimensional fluorescence spectrum data. Among them, the oil-paper sample undergoes an electrical aging experiment, and the three-dimensional fluorescence spectrum data of the oil-paper sample after partial discharge is collected for measurement; Step S2: Perform parallel factor analysis on the three-dimensional fluorescence spectrum data to extract the characteristic parameters of the fluorescence components. Among them, a parallel factor analysis model is established based on the three-dimensional fluorescence spectrum data, and the characteristic parameters of the fluorescence components in the parallel factor analysis model are identified and extracted; Step S3: Establish an electrical aging stage identification model based on the characteristic parameters. Among them, cluster analysis is performed on the characteristic parameters based on the K clustering algorithm, and an electrical aging stage identification model is built; Step S4: Input the characteristic parameters into the electrical aging stage identification model and output the electrical aging stage of the oil-paper insulation. Among them, new characteristic parameters are input into the electrical aging stage identification model, and the determination result of the electrical aging stage of the oil-paper insulation is output.

2. The method for identifying the electrical aging stage of oil-paper insulation according to claim 1, wherein Preferably, in step S1, the form of partial discharge is needle-plate discharge, and the distance between the tip of the needle electrode and the plate electrode is 2 mm.

3. The method for identifying the electrical aging stage of oil-paper insulation according to claim 2, wherein An oil-paper sample with a diameter of 80 mm and a thickness of 1 mm is placed between the needle electrode and the plate electrode. The material of the plate electrode is brass, with a diameter of 75 mm and a thickness of 15 mm; the material of the needle electrode is pure tungsten, and its curvature radius is 50 μm.

4. A method for identifying the electrical aging stage of oil-paper insulation according to claim 1, characterized in that In step S1, the step-up voltage method is used to determine the partial discharge inception voltage PDIV. During the electrical aging process, the positive and negative voltages are maintained at 1.5 times the inception voltage PDIV. Insulating oil is taken once per hour until the insulating paper is broken down. During the electrical aging process, a three-way valve and an oil pump are used to take oil without power interruption.

5. A method for identifying the electrical aging stage of oil-paper insulation according to claim 1, characterized in that In step S1, when collecting the three-dimensional fluorescence spectrum data, the slit 1 of the steady-state transient fluorescence spectrometer is set to 5 nm, and the slit 2 is set to 0.3 nm; the excitation band range is set to 330 - 450 nm, and the emission band range is set to 350 - 750 nm; the excitation band scanning interval is set to 5 nm, and the emission band scanning interval is set to 2 nm; the spectrometer scanning dwell time is set to 0.02 s.

6. The method for identifying the electrical aging stage of oil-paper insulation according to claim 5, characterized in that, The characteristic parameters of the fluorescence components include excitation loading, emission loading, and relative fluorescence intensity.

7. A method for identifying the electrical aging stage of oil-paper insulation according to claim 6, characterized in that, Step S2 includes Step S21: Preprocess the collected three-dimensional fluorescence spectrum data. The preprocessing includes baseline correction, smoothing processing, and normalization; Step S22: Select the number of components R by comparing the core consistency, residual value, and model interpretation rate of the parallel factor analysis model; Step S23: Iterative decomposition: Initialize the feature vectors: the excitation wavelength feature vector a k , the emission wavelength feature vector b k and the sample pattern feature vector c k . Take the least squares error of the parallel factor analysis model as the objective function, and use the alternating least squares method ALS for iterative decomposition. By alternately optimizing the feature vectors, gradually approach the minimum value of the objective function. When the objective function reaches a stable state, the iteration stops; Step S24: After decomposition, obtain R sets of excitation wavelength eigenvectors a k , emission wavelength eigenvectors b k and sample pattern eigenvectors c k .

8. A method for identifying the electrical aging stage of oil-paper insulation according to claim 7, characterized in that, Step S3 includes: Step S31: Initialization of cluster centers: Select the comprehensive feature vectors (a k , b k , c k ) of different components k as the initial cluster centers, so that the cluster centers represent different feature distributions of the data; Step S32: Sample distance calculation: During the process of iteratively allocating samples, the Euclidean distance formula is used to calculate the distance between the sample and the cluster center. For sample i and cluster center j, their comprehensive feature vectors are respectively (a i , b i , c i ) and (a j , b j , c j ), and the corresponding Euclidean distance formula is: ， Among them, represents the distance of the excitation wavelength eigenvector, represents the distance of the emission wavelength eigenvector, represents the distance of the sample pattern eigenvector; Step S33: Update of the cluster center: Recalculate the new cluster center according to the feature vectors of all samples within the same cluster. For cluster j, the new cluster center is expressed as: ， Among them, is the sample set in cluster j, a i , b i , c i are the characteristic parameters of sample i Step S34: Evaluation and analysis of the electrical aging stage identification model: By comparing the feature vectors in different clusters, reveal the characteristic differences of the oil-paper insulation at different electrical aging stages.

9. A method for identifying the stage of electrical aging of oil-paper insulation according to claim 1, characterized in that In step S4, the comprehensive evaluation parameter E consists of the eigenvectors a k , b k , c k and the calculation formula is: ， Wherein: a k , b k and c k respectively represent the excitation wavelength, emission wavelength, and sample mode feature vector; a min , b min and c min respectively represent the minimum values of a k , b k and c k ; a max , b max and c max respectively represent the maximum values of a k , b k and c k ; , and are weight coefficients, and by calculating the comprehensive evaluation parameter E, the determination result of the oil-paper insulation electrical aging stage is output.

10. A method for identifying the stage of oil-paper insulation electrical aging according to claim 9, characterized in that, When 0.1 ≤ E < 0.3, output the electrical aging stage of the oil-paper insulation as the early aging stage; when 0.3 ≤ E < 0.6, output the electrical aging stage of the oil-paper insulation as the middle aging stage; when 0.6 ≤ E < 0.9, output the electrical aging stage of the oil-paper insulation as the late aging stage.

Images