A spectral model transfer method and system for quantitatively evaluating the degree of polymerization of insulating paper
Through improved multivariate scattering correction method and model update technology, the inappropriate problem of insulating paper polymerization evaluation model caused by differences between different spectral instruments in different brands is solved, and the accurate quantity evaluation of insulating paper polymerization is achieved, which improves the accuracy of the model and applicability.
Patent Information
- Application Number
- CN202111313279.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-11-08
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2041-11-08
AI Technical Summary
The hardware differences between different brands of near-infrared spectral instruments lead to the inapplicability of insulating paper polymerization evaluation models, limiting the large-scale application of near-infrared spectral analysis technology in the diagnosis of insulating paper aging.
The improved multivariate scattering correction method combined with model updates is used to preprocess the spectral data through Savitzky-Golay convolution method and matrix standardization method, and the training samples based on the PLS model are constructed, the optimal number of PLS factors is obtained and the model is trained, and the model is trained. Evaluation model for quantitative evaluation of the polymerization degree of insulating paper.
It effectively compensates for the hardware differences between master-slave spectral instruments, reduces measurement noise, improves the accuracy and applicability of the model, and realizes accurate quantity evaluation of the polymerization degree of insulating paper.
Smart Images

Figure CN114036456B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of quantitative evaluation, and relates to a near infrared spectroscopy model transmission method, and specifically to a method for transmitting a main instrument spectral calibration model to a slave instrument in quantitative evaluation of the degree of polymerization of insulating paper. Background Art
[0002] Near infrared spectroscopy is an electromagnetic wave between mid-infrared and visible light. Its wavelength range overlaps with the frequency-doubled and frequency-combined absorption peaks of hydrogen-containing groups in substances. Therefore, by scanning the near infrared spectrum of a substance, information about the hydrogen-containing groups of the substance can be obtained, and then an evaluation model can be obtained through analytical methods such as chemometrics to analyze the content or characteristics of a certain component of the substance. Because of its rapid and non-destructive detection characteristics, near infrared spectroscopy analysis technology has been widely used in the fields of food, agriculture, medicine, and petrochemicals in China in recent years.
[0003] Insulating paper is an important insulating material in power equipment, and its aging degree determines the insulation life of the entire transformer. At present, the main characterization indicator of the aging degree of insulating paper is the degree of polymerization, which is often determined by the viscosity method. This method requires sampling and testing, has a long detection cycle, and requires the use of hazardous chemical reagents such as acetone, which brings many inconveniences to researchers and maintenance personnel. The article "Research on Modeling Methods for Quantitative Evaluation of the Degree of Polymerization of Insulating Paper Using Near-infrared Spectroscopy" (Proceedings of the Chinese Society of Electrical Engineering, 2019, 39, Supplement) uses near-infrared spectroscopy to establish a quantitative evaluation model for the degree of polymerization of insulating paper. The model effect verifies the feasibility of quantitative evaluation of the degree of polymerization of insulating paper using near-infrared spectroscopy.
[0004] The near-infrared spectrum of insulating paper needs to be obtained by near-infrared spectrometer. However, in practical applications, there may still be differences in spectrometer hardware, especially between instruments of different brands, which will lead to the inapplicability of the evaluation model. Therefore, the difference between the calibration instrument (master instrument) and the new instrument (slave instrument) is the bottleneck restricting the large-scale application of near-infrared spectroscopy analysis technology in insulating paper aging diagnosis. Summary of the invention
[0005] The purpose of the present invention is to propose a model transfer method and system based on improved multivariate scattering correction combined with model updating, so as to solve the problem that the evaluation model may not be applicable due to differences in spectrometer hardware, especially between instruments of different brands.
[0006] In order to achieve the above technical objectives, the present application provides a spectral model transfer method for quantitatively evaluating the degree of polymerization of insulating paper, comprising the following steps:
[0007] Collecting first spectral data and second spectral data of an insulating paper sample, performing interpolation processing and multivariate scattering correction on the first spectral data and the second spectral data, and constructing an initial sample data set;
[0008] Based on the Savitzky-Golay convolution method and matrix normalization method, the initial sample data set is preprocessed, and the training sample based on the PLS model is constructed to obtain the optimal number of PLS factors of the PLS model;
[0009] Based on the PLS model corresponding to the optimal PLS factor number, the initial sample data set is trained to obtain the first model parameter and the second model parameter;
[0010] Based on the PLS model, an evaluation model for quantitatively evaluating the polymerization degree of insulating paper is constructed through the optimal PLS factor number, the first model parameter and the second model parameter.
[0011] Preferably, in the process of constructing the initial sample data set, linear interpolation processing is performed on the first spectral data and the second spectral data based on the abscissa of the first spectral data;
[0012] The linear interpolation formula is:
[0013]
[0014] Among them, x 0 , x 1 is the endpoint of the interpolation interval, f() is the absorbance of the second spectral data at the interpolation endpoint, p(x) is the linear fitting function of the interpolation interval, and the abscissa of the second spectral data is replaced by the abscissa of the second spectral data and substituted into p(x).
[0015] Preferably, after interpolation processing is performed on the first spectral data and the second spectral data, the method for multivariate scattering correction includes the following steps:
[0016] extracting the mean value of the first spectrum data as ideal spectrum data for multivariate scattering correction;
[0017] Based on the first spectrum data, the second spectrum data, and the ideal spectrum data, a univariate linear regression process is performed to obtain a baseline translation amount and an offset amount of the insulating paper sample;
[0018] Based on the baseline translation and offset, multivariate scattering correction is performed on the first spectral data and the second spectral data to construct an initial sample data set.
[0019] Preferably, in the process of multivariate scattering correction, the first spectral data and the second spectral data are subtracted from the baseline shift and divided by the offset to obtain a corrected spectrum, and an initial sample data set is constructed based on the corrected spectrum.
[0020] Preferably, based on the initial sample data set, the Savitzky-Golay convolution method is used for smoothing and denoising, the window width and the fitting order are set, and the initial sample data set is subjected to windowing processing, wherein the windowing processing method is:
[0021]
[0022] Where m is the window width, P s is the smoothing coefficient corresponding to the sth data point, P S =X′ l+s / X l+s , where X′ l+s is the spectral absorbance fitting value of the l+sth point, X l+s is the true value of the spectral absorbance at the l+sth point, X′ l+s It is obtained by fitting the sample spectrum through the least squares method.
[0023] Preferably, the absolute difference in absorbance between the first spectral data and the second spectral data is eliminated by using a matrix standardization method, wherein the matrix standardization method is:
[0024]
[0025] Among them, S′ is the standardized spectrum, μ is the spectrum mean, and σ is the spectrum standard deviation.
[0026] Preferably, the first spectral data and the second spectral data after the preprocessing operation are integrated, and a leave-one-out cross-validation method is used to obtain the difference between the predicted value and the true value, and to screen outliers;
[0027] Arrange the first spectral data without abnormality after verification from large to small according to the degree of polymerization, and define the samples ranked 5n-2 as the verification set, and the remaining samples as the training set. Use the second spectral data as the training set;
[0028] By iteratively calculating the principal component factor ncomp, the optimal number of principal components for PLS prediction is determined, and the optimal number of principal components for PLS prediction is used as the optimal number of PLS factors.
[0029] Preferably, the method for determining the optimal number of principal components for PLS prediction is:
[0030]
[0031] Among them, RMSE is the root mean square error of the corresponding PLS model, DP rel is the actual value of the sample aggregation degree, DP forecast is the predicted value of sample aggregation degree.
[0032] A spectral model transfer system for insulation paper polymerization degree, comprising:
[0033] A data acquisition module, used for acquiring first spectrum data and second spectrum data of the insulating paper sample;
[0034] A first data processing module is used to perform interpolation processing and multivariate scattering correction on the first spectral data and the second spectral data to construct an initial sample data set;
[0035] The second data processing module is used to pre-process the initial sample data set based on the Savitzky-Golay convolution method and the matrix normalization method, construct a training sample based on the PLS model, and obtain the optimal PLS factor number of the PLS model;
[0036] A third data processing module is used to train the initial sample data set based on the PLS model corresponding to the optimal PLS factor number to obtain the first model parameter and the second model parameter;
[0037] A data evaluation module, for constructing an evaluation model for quantitatively evaluating the degree of polymerization of insulating paper based on a PLS model through an optimal number of PLS factors, a first model parameter, and a second model parameter;
[0038] A database module, used for storing system data of the spectral model delivery system;
[0039] The database logic control module is used to realize the system functions of the spectral model transmission system by controlling the database module to exchange data with other functional modules of the spectral model transmission system.
[0040] Preferably, the spectral model delivery system further comprises a computer program for implementing the quantitative evaluation method.
[0041] The present invention discloses the following technical effects:
[0042] 1. The improved multivariate scattering correction method proposed in the present invention uses the average spectrum of the master spectrometer as a benchmark, which to a certain extent makes up for the hardware differences between the master and slave spectrometers, making it possible to update the model of the master and slave spectrometers.
[0043] 2. The present invention takes into account the environmental differences and machine differences when the master and slave spectrometers collect spectra, and proposes a Savitzky-Golay convolution method and a matrix normalization method, which can effectively eliminate noise in the measurement process and reduce the absolute difference in absorbance between the master and slave spectrometers.
[0044] 3. The leave-one-out cross-validation method adopted in the present invention can effectively eliminate abnormal samples and improve the quality of the training set. At the same time, the method can select the optimal number of PLS principal components, which can improve the accuracy of the model to a certain extent.
[0045] 4. The PLS algorithm based on model updating proposed in the present invention is applied to the model transfer of the quantitative analysis model of the degree of polymerization of insulating paper, and can achieve better results than the traditional DS, PDS and Shenk's. BRIEF DESCRIPTION OF THE DRAWINGS
[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0047] Figure 1 This is a flow chart of the model transfer method of the present invention;
[0048] Figure 2 The spectrum collected by the host and the slave provided in the embodiment of the present invention;
[0049] Figure 3 Improve the spectrum image after multi-scattering correction for the slave spectrum;
[0050] Figure 4 Scatter plot of the correlation between the predicted value and the true value of the spectral data of the slave machine that did not participate in the modeling;
[0051] Figure 5 The figure is a comparison between the model transfer method described in the present invention and the traditional method. DETAILED DESCRIPTION
[0052] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.
[0053] like Figure 1-5 As shown, the present invention provides a spectral model transfer method for quantitatively evaluating the degree of polymerization of insulating paper, comprising the following steps:
[0054] Collecting first spectral data and second spectral data of an insulating paper sample, performing interpolation processing and multivariate scattering correction on the first spectral data and the second spectral data, and constructing an initial sample data set;
[0055] Based on the Savitzky-Golay convolution method and matrix normalization method, the initial sample data set is preprocessed, and the training sample based on the PLS model is constructed to obtain the optimal number of PLS factors of the PLS model;
[0056] Based on the PLS model corresponding to the optimal PLS factor number, the initial sample data set is trained to obtain the first model parameter and the second model parameter;
[0057] Based on the PLS model, an evaluation model for quantitatively evaluating the polymerization degree of insulating paper is constructed through the optimal PLS factor number, the first model parameter and the second model parameter.
[0058] Further preferably, in the process of constructing the initial sample data set, based on the abscissa of the first spectral data, linear interpolation processing is performed on the first spectral data and the second spectral data;
[0059] The linear interpolation formula is:
[0060]
[0061] Among them, x 0 , x 1 is the endpoint of the interpolation interval, f( ) is the absorbance of the second spectral data at the interpolation endpoint, p(x) is the linear fitting function of the interpolation interval, and the abscissa of the second spectral data is replaced by the abscissa of the second spectral data and substituted into p(x).
[0062] Further preferably, after interpolation processing is performed on the first spectral data and the second spectral data, the method for multivariate scattering correction includes the following steps:
[0063] extracting the mean value of the first spectrum data as ideal spectrum data for multivariate scattering correction;
[0064] Based on the first spectrum data, the second spectrum data, and the ideal spectrum data, a univariate linear regression process is performed to obtain a baseline translation amount and an offset amount of the insulating paper sample;
[0065] Based on the baseline translation and offset, multivariate scattering correction is performed on the first spectral data and the second spectral data to construct an initial sample data set.
[0066] Further preferably, in the process of multivariate scattering correction, the first spectral data and the second spectral data are subtracted from the baseline shift and divided by the offset to obtain a corrected spectrum, and an initial sample data set is constructed based on the corrected spectrum.
[0067] Further preferably, based on the initial sample data set, the Savitzky-Golay convolution method is used for smoothing and denoising, the window width and the fitting order are set, and the initial sample data set is subjected to windowing processing, wherein the windowing processing method is:
[0068]
[0069] Where m is the window width, P s is the smoothing coefficient corresponding to the sth data point, P S =X′ l+s / X l+s , where X′ l+s is the spectral absorbance fitting value of the l+sth point, X l+s is the true value of the spectral absorbance at the l+sth point, X′ l+s It is obtained by fitting the sample spectrum through the least squares method.
[0070] Further preferably, the absolute difference in absorbance between the first spectral data and the second spectral data is eliminated by using a matrix standardization method, wherein the matrix standardization method is:
[0071]
[0072] Among them, S′ is the standardized spectrum, μ is the spectrum mean, and σ is the spectrum standard deviation.
[0073] Further preferably, the first spectral data and the second spectral data after the preprocessing operation are integrated, and a leave-one-out cross-validation method is used to obtain the difference between the predicted value and the true value, and to screen outliers;
[0074] Arrange the first spectral data without abnormality after verification from large to small according to the degree of polymerization, and define the samples ranked 5n-2 as the verification set, and the remaining samples as the training set. Use the second spectral data as the training set;
[0075] The optimal number of principal components for PLS prediction is determined by iterative calculation of the principal component factor ncomp, and the optimal number of principal components for PLS prediction is taken as the optimal number of PLS factors.
[0076] Further preferably, the method for determining the optimal number of principal components for PLS prediction is:
[0077]
[0078] Among them, RMSE is the root mean square error of the corresponding PLS model, DP rel is the actual value of the sample aggregation degree, DP forecast is the predicted value of sample aggregation degree.
[0079] A spectral model transfer system for insulation paper polymerization degree, comprising:
[0080] A data acquisition module, used for acquiring first spectrum data and second spectrum data of the insulating paper sample;
[0081] A first data processing module is used to perform interpolation processing and multivariate scattering correction on the first spectral data and the second spectral data to construct an initial sample data set;
[0082] The second data processing module is used to pre-process the initial sample data set based on the Savitzky-Golay convolution method and the matrix normalization method, construct a training sample based on the PLS model, and obtain the optimal PLS factor number of the PLS model;
[0083] A third data processing module is used to train the initial sample data set based on the PLS model corresponding to the optimal PLS factor number to obtain the first model parameter and the second model parameter;
[0084] A data evaluation module, for constructing an evaluation model for quantitatively evaluating the degree of polymerization of insulating paper based on a PLS model through an optimal number of PLS factors, a first model parameter, and a second model parameter;
[0085] A database module, used for storing system data of the spectral model delivery system;
[0086] The database logic control module is used to realize the system functions of the spectral model transmission system by controlling the database module to exchange data with other functional modules of the spectral model transmission system.
[0087] Further preferably, the spectral model delivery system further includes a computer program for implementing the quantitative evaluation method.
[0088] Example 1: The present invention discloses a near-infrared spectroscopy model transfer method for quantitative evaluation of the degree of polymerization of insulating paper, which belongs to the field of spectral detection and spectrometer model transfer. The method is based on an improved model transfer method of multivariate scattering correction combined with model updating, and the specific contents include: obtaining the spectral data of the master instrument standard sample and a small amount of sample data from the slave instrument and performing interpolation processing, taking the mean of the master instrument standard sample spectrum as the "ideal spectrum" for multivariate scattering correction, and performing spectral multivariate scattering correction on the master instrument standard sample spectrum and the slave instrument; using the Savitzky-Golay convolution method and matrix standardization to pre-process the spectra of the master instrument and the slave instrument; mixing the master instrument standard sample spectrum with a small amount of slave instrument spectra as training samples, using the leave-one-out cross-validation method to obtain the optimal parameters of the PLS model, and determining the number of training sets and test sets; using the MUPLS method after finding the optimal parameters to establish the final model. Compared with the traditional method, the method proposed by the present invention can obtain the best effect with only a small number of slave instrument spectra, and realizes model transfer between different spectrometers.
[0089] Specifically, the present invention provides a near infrared spectroscopy model transfer method for quantitative evaluation of the degree of polymerization of insulating paper, comprising the following steps:
[0090] S1. Obtain the spectral data of the standard insulating paper sample of the main instrument and the spectral data of a small amount of insulating paper samples from the instrument, and perform multivariate scattering correction on the spectrum of the standard sample of the main instrument and the spectrum of the slave instrument;
[0091] S2. Spectral preprocessing of the master and slave instruments using Savitzky-Golay convolution and matrix normalization;
[0092] S3. Mix the master instrument standard sample spectrum with a small number of slave instrument spectra as training samples, and use the leave-one-out cross-validation PLS model to establish the optimal training set and test set; and determine the optimal number of PLS factors.
[0093] S4. Use the MUPLS method after finding the optimal parameters to establish the final model.
[0094] S5. Input the spectral data of other samples tested by the instrument into the above final model to obtain the prediction results.
[0095] S101 Specifically, step S1 includes: this implementation case uses a portable near-infrared spectrometer produced by Bidatech as the host A, with a wavelength range of 900-2500nm, and the slave uses a portable near-infrared spectrometer produced by Shanghai Maishe Company, with a wavelength range of 800-2500nm. 45 DLZ cable paper samples with different polymerization degree gradients were collected. When the slave was collecting, 8 paper samples were taken as training sets, and the rest were used as validation sets for model evaluation. The near-infrared spectra scanned by the host and slave instruments are linearly interpolated according to the horizontal coordinates of the spectrum data of the master instrument to obtain the spectrum data of the master and slave instruments with the same horizontal coordinates, such as Figure 2 As shown;
[0096] S102 further uses the average value of the spectrum of the standard sample of the main instrument as the "ideal spectrum" for multivariate scattering correction. The formula is: S Former is the main instrument spectral data, is the average of the spectrum of the main instrument, i.e., the "ideal spectrum", n is the number of spectral data, and the spectra of the two paper samples collected by the slave before and after correction are shown in Figure 2 .
[0097] S103 performs a univariate linear regression on the master instrument standard spectrum and the slave instrument and average spectrum to solve the least squares problem to obtain the baseline translation and offset of each sample. The formula is: Where S Latter is the spectrum from the instrument, k i With b i is the desired least squares coefficient.
[0098] S104 calibrates the spectrum of each sample by subtracting the obtained baseline shift and dividing by the offset to obtain a calibrated spectrum; S i The original spectrum of the master or slave instrument, S MSC is the spectrum after multivariate scattering correction.
[0099] S105 Further, the linear interpolation formula is:
[0100]
[0101] Among them, x 0 , x 1 is the endpoint of the interpolation interval, f( ) is the absorbance of the slave instrument at the interpolation endpoint, p(x) is the first-order fitting function of the interpolation interval, and the horizontal coordinate of the slave instrument is replaced by the horizontal coordinate of the main instrument and then p(x) is entered.
[0102] S201 Further, step S2 includes using Savitzky-Golay convolution method to smooth and denoise; setting the window width to 5 and the fitting order to 1, and then performing windowing processing on the spectral data S:
[0103]
[0104] Among them, m is the window width, P S is the smoothing coefficient corresponding to the sth data point, which is calculated by the following formula: S =X′ l+s / X l+s , where X′ l+s is the spectral absorbance fitting value of the l+sth point, X l+s is the true value of the spectral absorbance at the l+sth point. l+s It is obtained by fitting the sample spectrum through the least squares method.
[0105] S202, using the matrix normalization method to eliminate the absolute difference in absorbance between the master and slave instruments;
[0106]
[0107] Among them, S′ is the standardized spectrum, μ is the spectrum mean, and σ is the spectrum standard deviation.
[0108] S301, further, the step S3 includes: integrating the spectral data of the master instrument and the slave instrument after the preprocessing operation, using the leave-one-out cross-validation method to obtain the difference between the predicted value and the true value of each sample, and screening outliers.
[0109] S302, preferably, a prediction deviation greater than 150 is considered an outlier, and the paper sample selected this time has no outlier.
[0110] S303, arrange the host samples without abnormalities after verification from large to small according to the aggregation degree, set the samples ranked 5n-2 as the verification set, and the remaining samples as the training set. Randomly select 8 slave samples as the training set.
[0111] S304, preferably, by iteratively calculating the principal component factor ncomp, ncomp is iterated from 3 to 10 plus 1, to determine the PLS predicted response load ncomp=8, that is, the number of principal components is 8.
[0112] S401, further, the step S4 includes: using the PLS model after determining the optimal number of principal components, training the selected master and slave instrument training set samples. The training here is to train the samples obtained from the master and slave instruments together. The final model parameter b is obtained i with a i .
[0113] S501, further, the step S5 includes: performing MSC, Savitzky-Golay convolution smoothing and standardization operations on the non-training set sample spectra obtained from the instrument to calculate:
[0114] DP forecast =a i *S latter +b i
[0115] DP forecast is the degree of polymerization predicted from the instrumental sample, S latter For the spectrum from the instrument.
[0116] S502, further, the evaluation index of model prediction performance, determination coefficient R 2 The calculation formula of root mean square error RMSE is:
[0117]
[0118]
[0119] Among them, DP rel is the actual value of the sample aggregation degree, DP forecast is the predicted value of sample aggregation degree.
[0120] S503 is a slave sample used for verification, and its DP forecast With DP rel The distribution of Figure 3 .
[0121] S504 Specifically, R 2 The RMSE data are compared with the traditional DS, PDS, and Shenk's methods. Figure 4 .
[0122] It should be noted that similar numbers and letters represent similar items in the following figures. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are only used to distinguish the description and are not to be understood as indicating or implying relative importance.
[0123] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, those skilled in the art should understand that any person skilled in the art can still modify the technical solutions described in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or perform equivalent replacements on some of the technical features thereof; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. They should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims
1. A spectral model transfer method for quantitatively evaluating the degree of polymerization of insulating paper, characterized in that: The following steps are involved: Collecting first spectral data and second spectral data of an insulating paper sample, performing interpolation processing and multivariate scattering correction on the first spectral data and the second spectral data, and constructing an initial sample data set; Based on the Savitzky-Golay convolution method and the matrix normalization method, the initial sample data set is preprocessed, a training sample based on the PLS model is constructed, and the optimal number of PLS factors of the PLS model is obtained; Based on the PLS model corresponding to the optimal PLS factor number, training the initial sample data set to obtain a first model parameter and a second model parameter; Based on the PLS model, an evaluation model for quantitatively evaluating the degree of polymerization of insulating paper is constructed by using the optimal number of PLS factors, the first model parameters, and the second model parameters; In the process of constructing the initial sample data set, based on the abscissa of the first spectral data, linear interpolation processing is performed on the first spectral data and the second spectral data; The linear interpolation formula is: Wherein, x2 and x1 are the endpoints of the interpolation interval, f() is the absorbance of the second spectral data at the interpolation endpoint, and p(x) is the first-order fitting function of the interpolation interval to be sought. The abscissa of the second spectral data is replaced by the abscissa of the second spectral data and substituted into p(x); After interpolation processing is performed on the first spectral data and the second spectral data, the multivariate scattering correction method includes the following steps: Extracting the mean of the first spectral data as the ideal spectral data for multivariate scattering correction; Based on the first spectral data, the second spectral data, and the ideal spectral data, a univariate linear regression process is performed to obtain a baseline translation amount and an offset amount of the insulating paper sample; Based on the baseline translation and the offset, multivariate scattering correction is performed on the first spectral data and the second spectral data to construct the initial sample data set.
2. A spectral model transfer method for quantitatively evaluating the degree of polymerization of insulating paper according to claim 1, characterized in that: In the process of multivariate scattering correction, the first spectral data and the second spectral data are subtracted from the baseline shift and divided by the offset to obtain a corrected spectrum, and the initial sample data set is constructed based on the corrected spectrum.
3. The spectral model transfer method for quantitatively evaluating the degree of polymerization of insulating paper according to claim 1, characterized in that: Based on the initial sample data set, the Savitzky-Golay convolution method is used for smoothing and denoising, the window width and the fitting order are set, and the initial sample data set is subjected to windowing processing, wherein the windowing processing method is: Where m is the window width, Ps is the smoothing coefficient corresponding to the sth data point, and P s =X1' +s / X 1+s , where X1' +s is the spectral absorbance fitting value of the l+sth point, X 1+s is the true value of the spectral absorbance at the l+sth point, X1' +s It is obtained by fitting the sample spectrum through the least squares method.
4. A spectral model transfer method for quantitatively evaluating the degree of polymerization of insulating paper according to claim 3, characterized in that: By using the matrix standardization method, the absolute difference in absorbance between the first spectral data and the second spectral data is eliminated, wherein the matrix standardization method is: Among them, S′ is the standardized spectrum, μ is the spectrum mean, and σ is the spectrum standard deviation.
5. A spectral model transfer method for quantitatively evaluating the degree of polymerization of insulating paper according to claim 4, characterized in that: Performing data integration on the first spectral data and the second spectral data after the preprocessing operation, using leave-one-out cross-validation to obtain the difference between the predicted value and the true value, and screening outliers; Arrange the first spectral data without abnormality after verification from large to small according to the degree of polymerization, define the samples ranked 5n-2 as the verification set, and the remaining samples as the training set; and use the second spectral data as the training set; The optimal number of principal components predicted by PLS is determined by iterative calculation of the principal component factor ncomp, and the optimal number of principal components predicted by PLS is used as the optimal number of PLS factors.
6. A spectral model transfer method for quantitatively evaluating the degree of polymerization of insulating paper according to claim 5, characterized in that: The method to determine the optimal number of principal components for PLS prediction is: Among them, RMSE is the root mean square error corresponding to the PLS model, DP rel is the actual value of the sample aggregation degree, DP forecast is the predicted value of sample aggregation degree.
7. A spectral model transfer system for quantitatively evaluating the degree of polymerization of insulating paper, used to execute the spectral model transfer method for quantitatively evaluating the degree of polymerization of insulating paper according to any one of claims 1 to 6, characterized in that: include: A data acquisition module, used for acquiring first spectrum data and second spectrum data of the insulating paper sample; A first data processing module, used for performing interpolation processing and multivariate scattering correction on the first spectral data and the second spectral data to construct an initial sample data set; The second data processing module is used to pre-process the initial sample data set based on the Savitzky-Golay convolution method and the matrix normalization method, construct a training sample based on the PLS model, and obtain the optimal PLS factor number of the PLS model; A third data processing module is used to train the initial sample data set based on the PLS model corresponding to the optimal PLS factor number to obtain a first model parameter and a second model parameter; A data evaluation module, configured to construct an evaluation model for quantitatively evaluating the degree of polymerization of insulating paper based on the PLS model through the optimal number of PLS factors, the first model parameters, and the second model parameters; A database module, used for storing system data of the spectral model transmission system; The database logic control module is used to realize the system functions of the spectral model transmission system by controlling the database module to perform data interaction with other functional modules of the spectral model transmission system.
8. A spectral model transfer system for quantitatively evaluating the degree of polymerization of insulating paper according to claim 7, characterized in that: The spectral model transfer system further includes a computer program for implementing the spectral model transfer method according to any one of claims 1 to 6.
