Transformer oil chromatographic data analysis method, device, equipment and medium

Abstract

The present application belongs to the technical field of transformer oil chromatographic online monitoring, and discloses a transformer oil chromatographic data analysis method, device, equipment and medium, comprising: performing smoothing operation on spectrum data obtained by oil chromatographic online monitoring to obtain smoothed data; performing first-order derivation on the smoothed data to obtain zero-crossing points, and preliminarily determining chromatographic peak positions and sizes according to the zero-crossing points; correcting peak retention time offset of multi-component gas according to a linear regression model, the linear regression model being obtained according to pre-column pressure change fitting; and filtering out miscellaneous peaks by using a multi-factor weighted scoring model to obtain final chromatographic peak positions and sizes. The present application can improve the stability and reliability of oil chromatographic online monitoring analysis, and effectively reduce the risk of false positives and false negatives.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of transformer oil chromatography online monitoring, and particularly relates to a transformer oil chromatography data analysis method, device, equipment and medium. BACKGROUND

[0002] In the development of transformer potential fault online monitoring, the oil chromatography online monitoring technology is one of common means for transformer fault monitoring. By continuously monitoring the dissolved gas (H2, CO, CH4, CO2, C2H4, C2H6 and C2H2) in oil, the running state of the transformer equipment can be determined.

[0003] However, the long-term operation of the chromatographic column unit is affected by temperature, humidity and oil quality environment, which may cause chromatographic peak distortion or deviation, affect the stability and reliability of the oil chromatography online monitoring analysis result, and easily cause misjudgment. SUMMARY

[0004] The application aims to provide a transformer oil chromatography data analysis method, device, equipment and storage medium, which can improve the stability and reliability of oil chromatography online monitoring analysis and effectively reduce the false alarm and missed alarm risk.

[0005] The application discloses a transformer oil chromatography data analysis method, which comprises the following steps:

[0006] Smooth the spectrum data obtained by oil chromatography online monitoring to obtain smooth data;

[0007] First-derivate the smooth data to obtain zero-crossing points, and preliminarily determine the chromatographic peak position and size according to the zero-crossing points;

[0008] Correct the peak retention time deviation of multi-component gas according to a linear regression model, wherein the linear regression model is obtained by fitting according to the change of pre-column pressure;

[0009] Filter the impurity peaks by using a multi-factor weighted scoring model to obtain the final chromatographic peak position and size.

[0010] In some embodiments, the smooth spectrum data obtained by oil chromatography online monitoring comprises the following steps:

[0011] Smooth the spectrum data by using a Gaussian mixture model to obtain first smooth data;

[0012] Smooth the first smooth data by using mean filtering to obtain second smooth data, and set the second smooth data as the smooth data.

[0013] In some embodiments, the smooth spectrum data obtained by oil chromatography online monitoring comprises the following steps:

[0014] acquiring an effective narrowest peak width size in the chromatographic sampling;

[0015] setting a window width as 27% of the effective narrowest peak width size, performing mean filtering on the first smoothing data according to the window width to obtain the second smoothing data.

[0016] In some embodiments, the filtering out of the impurity peak by using the multi-factor weighted scoring model to obtain the final chromatographic peak position and size comprises:

[0017] For each chromatographic peak, the following operations are performed: when the convexity of the chromatographic peak is greater than or equal to a set threshold, the kurtosis, skewness, symmetry and half-peak width of the chromatographic peak are calculated and normalized, respectively, and then the weighted statistics are performed on the normalized values to obtain the peak shape score of the chromatographic peak; otherwise, the peak shape score of the chromatographic peak is set to 0.

[0018] The chromatographic peak position and size corresponding to the maximum peak shape score are set as the final chromatographic peak position and size.

[0019] In some embodiments, the weights corresponding to the kurtosis, skewness, symmetry and half-peak width of the chromatographic peak in the weighted statistics are obtained by training historical samples by using a random forest method.

[0020] In some embodiments, the peak retention time offset of the multi-component gas is corrected according to a linear regression model, comprising:

[0021] The peak retention time offset of each of ethylene, ethane and acetylene is corrected according to the linear regression model corresponding to each of them, respectively, wherein the linear regression model corresponding to each of ethylene, ethane and acetylene is fitted according to the historical data of the pre-column pressure conversion and based on carbon dioxide.

[0022] In some embodiments, the step of obtaining the linear regression model corresponding to the target component gas comprises:

[0023] According to the historical data of the pre-column pressure conversion, the peak vertex time increment of the pre-column pressure change of carbon dioxide and the target component gas is calculated;

[0024] Taking the peak vertex time increment of carbon dioxide as the independent variable and the peak vertex time increment of the target component gas as the dependent variable, the regression parameters are calculated by using the least square method to obtain the linear regression model corresponding to the target component gas.

[0025] The second aspect of the present application discloses a transformer oil chromatographic data analysis device, comprising:

[0026] The smoothing module is configured to perform smoothing operation on the spectrum data obtained by the oil chromatographic online monitoring to obtain smoothing data.

[0027] a first-order derivation module configured to perform first-order derivation on the smoothed data to obtain zero-crossing points, and preliminarily determine the positions and sizes of the chromatographic peaks according to the zero-crossing points;

[0028] a peak retention time offset correction module configured to correct the peak retention time offset of the multi-component gas according to a linear regression model, which is obtained by fitting according to the change in pre-column pressure;

[0029] a peak filtering module configured to filter out the impurity peaks by using a multi-factor weighted scoring model, and obtain the final positions and sizes of the chromatographic peaks.

[0030] The third aspect of the present application discloses an electronic device, comprising a memory storing executable program codes and a processor coupled with the memory; the processor invokes the executable program codes stored in the memory, and is configured to execute the transformer oil chromatographic data analysis method disclosed in the first aspect.

[0031] The fourth aspect of the present application discloses a computer readable storage medium storing a computer program, wherein the computer program enables a computer to execute the transformer oil chromatographic data analysis method disclosed in the first aspect.

[0032] The present application has the beneficial effect that, by smoothing the chromatogram data, the first-order derivation is performed on the smoothed data to find the zero-crossing points, the peak retention time offset of the multi-component gas is corrected according to the fitted linear regression model, and the multi-factor weighted scoring model is used to filter out the impurity peaks for each chromatographic peak meeting the peak retention time range, so as to determine the positions and sizes of the final chromatographic peaks. The stability and reliability of the online monitoring and analysis of the oil chromatogram can be improved, and the false alarm and missed alarm risks can be effectively reduced. BRIEF DESCRIPTION OF DRAWINGS

[0033] The drawings herein show the specific examples of the technical solutions of the present application, and constitute a part of the specification together with the specific embodiments, for explaining the technical solutions, principles and effects of the present application.

[0034] Unless specifically stated or defined otherwise, the same reference signs in different drawings represent the same or similar technical features, and different reference signs may also be used to represent the same or similar technical features.

[0035] Figure 1 is a flowchart of a transformer oil chromatographic data analysis method disclosed by an embodiment of the present application;

[0036] Figure 2 is a flowchart of obtaining a linear regression model according to an embodiment of the present application;

[0037] Figure 3 is a structural schematic diagram of a transformer oil chromatographic data analysis device according to an embodiment of the present application;

[0038] Figure 4 is a structural schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0039] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the description of the application herein can also have the meanings commonly understood by those of ordinary skill in the art to which the application pertains, except to the extent that the context in which used indicates otherwise. As used herein, the articles "a", "an" and "the" are intended to mean zero or one or more of the specified element or element. The articles "a" and "an" are used interchangeably in the disclosure to mean one or more of the associated noun(s) that it precedes, which can be dependent upon context. The term "and / or" used in the applications means one or all the open associated listed terms.

[0040] It should be noted that when an element is referred to as being "on" another element, it can be directly on the other element or intervening elements can also be present. In addition, when an element is referred to as being "connected" to another element, it can be directly connected to the other element or intervening elements can also be present. In addition, when an element is referred to as being "mounted on" another element, it can be directly mounted on the other element or intervening elements can also be present. When an element is referred to as being "provided on" another element, it can be directly provided on the other element or intervening elements can also be present.

[0041] Unless otherwise defined, "said", "the" (when used) refer to the technical features or technical contents previously mentioned or described in the corresponding position, which can be the same as or similar to the technical features or technical contents mentioned. In addition, the terms "comprise" and "have" and any variations thereof used herein are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units not listed, or optionally also includes other steps or units inherent to the process, method, product or device.

[0042] Due to the influence of temperature, humidity and oil quality environment during long-term operation of the chromatographic column unit, the chromatographic peak is deformed or offset, which can cause misjudgment and missed report of the oil chromatographic online monitoring analysis result. In order to improve the accuracy of the online monitoring analysis result, the original spectrum data is smoothed by using the Gaussian mixture model and the mean filtering in the present application; the smoothed data is differentiated to find the zero point, and the chromatographic peak position and size are preliminarily determined. According to the multi-component gas unary linear regression environmental model fitted by the pre-column pressure change, the peak retention time offset is corrected, and the multi-factor weighted scoring model composed of kurtosis, skewness, symmetry, convexity and half-peak width is used to filter out the impurity peak, so as to determine the final chromatographic peak position and size. The stability and reliability of real-time oil chromatographic online monitoring data analysis can be significantly improved, and the risk of false report and missed report can be effectively reduced.

[0043] The transformer oil chromatographic data analysis method disclosed in the embodiments of the present application can be realized by computer programming. The execution subject of the method can be an electronic device such as a computer, a notebook computer, a tablet computer, or a control chip embedded in the electronic device, and the present application does not limit this.

[0044] In order to facilitate the understanding of the present application, the specific embodiments of the present application will be described in more detail below with reference to the accompanying drawings.

[0045] As shown in the method shown in the figure, the method comprises the following steps: Figure 1

[0046] Step S100: smoothing the spectrum data obtained by oil chromatographic online monitoring to obtain smoothed data;

[0047] In the transformer oil chromatographic online monitoring, the spectrum data refers to the original detection signal data and the characteristic data derived based on the signal generated after the monitoring system separates and detects the dissolved component gas (such as hydrogen H2, methane CH4, ethane C2H6, ethylene C2H4, acetylene C2H2, carbon monoxide CO, carbon dioxide CO2, etc.) in the transformer oil by gas chromatographic analysis technology. These data not only contain the graphical information of the chromatogram, but also cover the core numerical information supporting qualitative and quantitative analysis, which is the key basis for judging whether there is a fault in the transformer.

[0048] After obtaining the spectrum data, first, the Gaussian mixture model is used to smooth the spectrum data to obtain first smoothed data; then the mean filtering is used to smooth the first smoothed data again to obtain second smoothed data, and the obtained second smoothed data is taken as the final smoothed data.

[0049] ​Specifically, the process of one smoothing using Gaussian mixture model is as follows: the calculation formula of normalized Gaussian kernel is used as weight to convolve the spectrum data of oil chromatogram in the window width range to achieve the effect of smoothing data.

[0050] ;

[0051] ;

[0052] wherein k represents the peak retention time, w represents the smallest half-peak width 20 of the gas phase, g(k) is a Gaussian kernel function, m is the window width taken, which is set to 61 in the embodiment, and N(k) is the normalized Gaussian kernel.

[0053] The convolution formula is wherein m is the window width taken, x(n) is the spectrum data, and y(n) is the one-smoothed data.

[0054] The process of smoothing the one-smoothed data again using mean filtering is as follows: the one-smoothed spectrum data is smoothed again using a suitable window width. In the embodiment, 27% of the effective narrowest peak width in the chromatogram sampling is taken as the window width, and then the one-smoothed data is filtered using mean filtering according to the window width to obtain the twice-smoothed data. The narrowest peak width is not pursued to be as narrow as possible, but needs to meet two effectiveness (the peak shape is complete and symmetrical, and the signal strength is sufficient), otherwise even if the peak width is extremely small, it also has no analysis value. In the embodiment, 27% of the effective narrowest peak width is taken as the window width, which locks the core area of the Gaussian peak with the highest signal strength and the most representative target component.

[0055] The one-smoothed data is twice-smoothed using moving average according to the following formula.

[0056]

[0057] wherein k is the window size, which is taken as 10 in the embodiment, x(n) is the one-smoothed data, and y(n) is the twice-smoothed data.

[0058] Step S200: first derivative is performed on the smoothed data to obtain the zero-crossing point, and the chromatographic peak position and size are preliminarily determined according to the zero-crossing point;

[0059] The zero-crossing point corresponds to the peak vertex, and the first derivative is performed on the twice-smoothed spectrum data according to the following formula to find the zero-crossing point and preliminarily determine the chromatographic peak position and size in the spectrum data.

[0060] ,

[0061] Wherein, x(n+Δt) is the data value of the n th point in the secondary smoothing data at the right side of the interval Δt position, x(n-Δt) is the data value of the n th point in the secondary smoothing data at the left side of the interval Δt position, Δt is the sampling interval, d(n) is the first derivative of the n th point, and the point where d(n) is 0 is the zero-crossing point.

[0062] By analyzing the zero-crossing point, the position and size of the chromatographic peak can be determined.

[0063] Step S300: correcting the peak retention time offset of each component gas according to a linear regression model, wherein the linear regression model is obtained by fitting according to the change of the pre-column pressure;

[0064] During long-term operation, changes in pre-column pressure caused by factors such as aging of the chromatographic column and temperature can cause data offset. Therefore, the coefficients between the peaks of different components are calculated according to the linear regression model to correct the peak retention time offset, so that the actual retention time of the target component is returned to the standard retention time range, thereby restoring the separation stability and qualitative accuracy of the chromatographic system. Specifically, correcting the peak retention time offset is to correct the system deviation rather than modifying the data itself, and the ultimate goal is to make the peak time of the component consistent with the standard value in subsequent analysis, and to ensure the reliability of the qualitative and quantitative results.

[0065] In this embodiment, the peak retention time offset of each of ethylene, ethane and acetylene is corrected according to the corresponding linear regression model of each of them, wherein the linear regression model of each of ethylene, ethane and acetylene is obtained by fitting according to the historical data of the change of the pre-column pressure with carbon dioxide as the reference. That is, for C2H4, C2H6 and C2H2, the linear regression model for C2H4, C2H6 and C2H2 is obtained by fitting the peak retention time offset trend of C2H4, C2H6 and C2H2 according to the historical data of the change of the pre-column pressure with CO2 as the reference.

[0066] As shown in Figure 2 The steps of obtaining the linear regression model corresponding to the target component gas (C2H4, C2H6, C2H2) include:

[0067] Step S310: calculating the peak top time increment of the pre-column pressure change of carbon dioxide and the target component gas according to the historical data of the change of the pre-column pressure;

[0068] Step S320: taking the peak top time increment of carbon dioxide as the independent variable and the peak top time increment of the target component gas as the dependent variable, and calculating the regression parameters by using the least square method to obtain the linear regression model corresponding to the target component gas.

[0069] Specifically, the data points of the linear regression model are the peak points of the chromatographic peaks of CO2, C2H4, C2H6 and C2H2 calibrated with a step of 0.01 in the range of 0.08-0.12 of the pre-column pressure.

[0070] The peak top time increments of the four-component gases CO2, C2H4, C2H6 and C2H2 with respect to the change of the pre-column pressure, i.e., (0.08→0.09), (0.09→0.1), (0.1→0.11), (0.11→0.12), (0.12→0.13) and (0.13→0.14), are respectively denoted as , , , , and is the independent variable, , , the dependent variable is calculated by the least square method to obtain a linear regression model for correcting the peak retention time shift. In theory, when the independent variable is 0, the predicted value of the dependent variable , , should be 0, and the influence of the intercept is small, so the intercept is ignored to obtain the following linear regression model.

[0071]

[0072]

[0073]

[0074] wherein, , , , is the peak top time increment of the pre-column pressure change.

[0075] Step S400: The multi-factor weighted scoring model is used to filter out the impurity peaks to obtain the final chromatographic peak position and size.

[0076] The multi-factor analysis includes: kurtosis (K), skewness (S), symmetry (S), convexity (C), and half-width at half-maximum (Hpw). Kurtosis and skewness are used to determine whether a sample in the chromatography exhibits normal statistical characteristics. Convexity supplements kurtosis, symmetry supplements skewness, and Hpw ensures the reliability and stability of the sample. The factor ranges for each component are derived from historical gas phase peak statistics. A kurtosis range of [-1.48, 3] indicates a suitable peak shape; a skewness range of [-1.8, 1.9] indicates a basically symmetrical peak shape; a symmetry range of [0.3, 3.4] indicates good symmetry; a convexity greater than or equal to 0.0001 indicates significant peak convexity; and a Hpw of [6, 60] indicates a suitable peak shape. If all the above factor conditions are met, the peak shape is considered a potentially high-quality gas phase peak; if some conditions are not met, the peak is considered a distorted, impure peak.

[0077] For each chromatographic peak, perform the following operations: when the convexity of the chromatographic peak is greater than or equal to the set threshold (e.g., 0.0001), calculate the peak intensity, skewness, symmetry, and half-width of the chromatographic peak and normalize them respectively. Then, perform weighted statistics on the normalized values ​​to obtain the peak shape score of the chromatographic peak; otherwise, set the peak shape score of the chromatographic peak to 0.

[0078] Then, the position and size of the chromatographic peak corresponding to the maximum peak shape score are set as the final chromatographic peak position and size.

[0079] Kurtosis is calculated using the following formula:

[0080] ,

[0081] Where n is the sample size from the peak start point to the peak end point. For the i-th sample, Let S be the sample mean from the peak start point to the peak end point, and let S be the sample standard deviation from the peak start point to the peak end point.

[0082] The ideal kurtosis value is 3, and its normalization formula is:

[0083] ,

[0084] in, It is -1.48. The value is 3.

[0085] The skewness is calculated using the following formula:

[0086]

[0087] Where n is the sample size from the peak start point to the peak end point. For the i-th sample, Let S be the sample mean from the peak start point to the peak end point, and let S be the sample standard deviation from the peak start point to the peak end point.

[0088] The ideal value of skewness is 0, and the normalization formula is .

[0089] The convexity calculation method is the ratio of the distance between the sample points from the peak starting point to the peak ending point to the connecting line of the peak starting point and the peak ending point to the length of the connecting line, and the value of the ratio is between 0 and 1.

[0090] The symmetry factor calculation method is to take the ratio of the peak width at 5% of the peak height to twice the front half peak width. The ideal value of the symmetry factor is 1, and the normalization formula is .

[0091] The half peak width calculation method is half peak width = peak right inflection point - peak left inflection point.

[0092] The half peak width normalization formula is , wherein Hpw min is 6, and Hpw max is 60.

[0093] The embodiment calculates the peak shape score of the chromatographic peak by using the following multi-factor weighted scoring model.

[0094] .

[0095] Each factor weight (such as 0.14, 0.21, 0.24, and 0.41) in the multi-factor weighted scoring model is obtained by training historical samples by using a random forest method.

[0096] The historical samples of the embodiment are shown in the following table:

[0097]

[0098] For the preliminarily determined chromatographic peak, after correcting the peak retention time offset, the peak shape score of each chromatographic peak meeting the peak retention time range is calculated, the peak shape score with the maximum value is taken as the final gas chromatographic peak, and the final chromatographic peak position and size are obtained.

[0099] To solve the problem that noise data can drown out and distort data characteristics, the embodiment smoothes the data once by using Gaussian filtering and smoothes the data twice by using mean filtering. The first derivative of the smoothed data is calculated to find the zero crossing point, the peak retention time offset is corrected according to the trained one-dimensional linear regression model, the peak is distinguished by combining the multi-factor weighted scoring model, and the position and size of the final peak are determined.

[0100] As shown in Figure 3 , based on the transformer oil chromatographic data analysis method, the embodiment of the present application discloses a transformer oil chromatographic data analysis device, which comprises:

[0101] The smoothing module 600 is configured to perform smoothing operation on the spectrum data obtained by the online monitoring of the oil spectrum to obtain smoothed data;

[0102] The first derivative module 610 is configured to perform first derivative operation on the smoothed data to obtain zero-crossing points, and preliminarily determine the positions and sizes of the chromatographic peaks according to the zero-crossing points;

[0103] The peak retention time offset correction module 620 is configured to correct the peak retention time offset of the multi-component gas according to a linear regression model, which is obtained by fitting according to the change of the pre-column pressure;

[0104] The impurity peak filtering module 630 is configured to filter out the impurity peaks by using a multi-factor weighted scoring model to obtain the final positions and sizes of the chromatographic peaks.

[0105] As shown in Figure 4 the embodiment of the present application discloses an electronic device, which comprises a memory 401 storing executable program codes and a processor 402 coupled with the memory 401;

[0106] The processor 402 calls the executable program codes stored in the memory 401 to execute the transformer oil chromatographic data analysis method described in the above embodiments.

[0107] The embodiment of the present application also discloses a computer readable storage medium storing a computer program, wherein the computer program enables the computer to execute the transformer oil chromatographic data analysis method described in the above embodiments.

[0108] The purpose of the above embodiments is to exemplarily reproduce and deduce the technical solutions of the present application, and to completely describe the technical solutions, purposes and effects of the present application, so as to make the public understand the disclosure of the present application more thoroughly and comprehensively, and not to limit the protection scope of the present application.

[0109] The above embodiments are not based on the exhaustive enumeration of the present application, and there can be many other unlisted embodiments. Any replacement and improvement made without violating the concept of the present application is within the protection scope of the present application.

Claims

1. A method of analyzing transformer oil chromatography data, characterized by, The method comprises the following steps: Smooth the spectrum data obtained by online monitoring of oil chromatography to obtain smooth data; First derivative is performed on the smooth data to obtain zero-crossing points, and the peak position and size of the chromatogram are preliminarily determined according to the zero-crossing points; The peak retention time offset of the multi-component gas is corrected according to a one-dimensional linear regression model, and the one-dimensional linear regression model is fitted according to the change of the pre-column pressure; A multi-factor weighted scoring model is used to filter out the impurity peaks to obtain the final peak position and size of the chromatogram; The multi-factor weighted scoring model is used to filter out the impurity peaks to obtain the final peak position and size of the chromatogram, which comprises the following steps: For each chromatographic peak, the following operations are performed: when the convexity of the chromatographic peak is greater than or equal to a set threshold, the kurtosis, skewness, symmetry and half-peak width of the chromatographic peak are calculated and normalized, and then the normalized values are weighted and counted to obtain the peak shape score of the chromatographic peak; otherwise, the peak shape score of the chromatographic peak is set to 0; The peak position and size corresponding to the maximum peak shape score are set as the final peak position and size of the chromatogram; The weights of the kurtosis, skewness, symmetry and half-peak width of the chromatographic peak in the weighted counting are obtained by training historical samples by using a random forest method; The peak retention time offset of the multi-component gas is corrected according to a one-dimensional linear regression model, which is fitted according to the change of the pre-column pressure; The peak retention time offset of ethylene, ethane and acetylene is respectively corrected according to the one-dimensional linear regression model corresponding to each of them, wherein the one-dimensional linear regression model corresponding to each of ethylene, ethane and acetylene is fitted according to the historical data of the change of the pre-column pressure of carbon dioxide as the reference; The step of obtaining the one-dimensional linear regression model corresponding to the target component gas comprises the following steps: According to the historical data of the change of the pre-column pressure, the peak vertex time increment of carbon dioxide and the target component gas is calculated; The least square method is used to calculate the regression parameters by taking the peak vertex time increment of carbon dioxide as the independent variable and the peak vertex time increment of the target component gas as the dependent variable, and the one-dimensional linear regression model corresponding to the target component gas is obtained.

2. The transformer oil chromatogram data analysis method of claim 1, wherein, The spectrum data obtained by online monitoring of oil chromatography is smoothed to obtain smooth data, which comprises the following steps: The spectrum data is smoothed by using a Gaussian mixture model to obtain first smooth data; The first smooth data is smoothed by using mean filtering to obtain second smooth data, and the second smooth data is set as the smooth data.

3. The transformer oil chromatogram data analysis method of claim 2, wherein, The first smooth data is smoothed by using mean filtering to obtain second smooth data, which comprises the following steps: The effective narrowest peak width size in chromatographic sampling is obtained; A window width of 27% of the effective narrowest peak width size is set, and the first smooth data is filtered by mean filtering according to the window width to obtain the second smooth data.

4. A transformer oil chromatogram data analysis device characterized by comprising: The method comprises the following steps: A smoothing module is configured to smooth the spectrum data obtained by online monitoring of oil chromatography to obtain smooth data; A first derivative module is configured to perform first derivative on the smooth data to obtain zero-crossing points, and preliminarily determine the peak position and size of the chromatogram according to the zero-crossing points; A peak retention time offset correction module is configured to correct the peak retention time offset of the multi-component gas according to a one-dimensional linear regression model, and the one-dimensional linear regression model is fitted according to the change of the pre-column pressure; The impurity peak filtering module is configured to filter out impurity peaks by using a multi-factor weighted scoring model to obtain final chromatographic peak positions and sizes. The multi-factor weighted scoring model is used to filter out impurity peaks and obtain final chromatographic peak positions and sizes, including: For each chromatographic peak, the following operations are performed: when the convexity of the chromatographic peak is greater than or equal to a set threshold, the kurtosis, skewness, symmetry and half-peak width of the chromatographic peak are calculated and normalized, respectively, and then the normalized values are weighted and counted to obtain a peak shape score of the chromatographic peak; otherwise, the peak shape score of the chromatographic peak is set to 0; The chromatographic peak position and size corresponding to the maximum peak shape score are set as the final chromatographic peak position and size; The weights corresponding to the kurtosis, skewness, symmetry and half-peak width of the chromatographic peak in the weighted counting are obtained by training historical samples using a random forest method; According to a linear regression model, the peak retention time offset of the multi-component gas is corrected, including: According to the respective linear regression model of ethylene, ethane and acetylene, the peak retention time offset of each is corrected, wherein the linear regression model of ethylene, ethane and acetylene is fitted according to the historical data of the pre-column pressure change with carbon dioxide as the reference; The step of obtaining the linear regression model corresponding to the target component gas includes: According to the historical data of the pre-column pressure change, the peak vertex time increment of carbon dioxide and the target component gas is calculated; Using the peak vertex time increment of carbon dioxide as the independent variable and the peak vertex time increment of the target component gas as the dependent variable, the regression parameters are calculated by using the least squares method to obtain the linear regression model corresponding to the target component gas.

5. An electronic device, comprising: The computer readable storage medium stores a computer program, wherein the computer program causes the computer to execute the transformer oil chromatographic data analysis method of any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, wherein the computer program causes the computer to execute the transformer oil chromatographic data analysis method of any one of claims 1-3.

Images