Transformer spectrogram baseline anomaly identification method based on similarity and Wasserstein distance

Abstract

A transformer spectrogram baseline anomaly identification method based on similarity and Wasserstein distance comprises the following steps: firstly, through the preprocessing steps of Gaussian filtering denoising, baseline correction, data cleaning, data integration, data transformation, data reduction and the like, the data quality is ensured; then, establishing a standard chromatogram library without abnormity, and extracting a baseline template; and calculating the overall similarity between the spectrogram to be detected and the standard spectrogram by using a Pearson correlation coefficient, and setting a similarity threshold to preliminarily screen an abnormal spectrogram. Meanwhile, the Wasserstein distance between the to-be-measured spectrogram base line and the standard base line template is calculated, and a distance threshold value is set so as to accurately identify small anomalies such as base line drift and the like. And determining whether the baseline anomaly exists in the spectrogram or not by comprehensively judging the results of the similarity and the Wasserstein distance. And finally, developing a graphical user interface, outputting an anomaly recognition result, including similarity, a Wasserstein distance value and an anomaly type, and providing an anomaly alarm function. The method improves the accuracy and reliability of anomaly detection, and is suitable for transformer maintenance and fault diagnosis.

Description

Technical Field

[0001] The invention belongs to the field of transformers, and in particular relates to a transformer spectrum baseline anomaly recognition method based on similarity and Wasserstein distance. Background Art

[0002] In the power system, the transformer is one of the key equipment, and its operating status directly affects the safety and stability of the entire power grid. Transformer oil chromatography analysis is an important fault diagnosis method. By analyzing the composition and content of dissolved gas in transformer oil, it can indirectly reflect the internal fault of the transformer. However, in practical applications, chromatograms are often affected by various factors, such as equipment accuracy, operating conditions, etc., resulting in abnormal phenomena such as baseline drift, ghost peaks or miscellaneous peaks. These abnormalities not only increase the difficulty of fault diagnosis, but also may mislead the diagnosis results.

[0003] Traditional chromatogram anomaly recognition methods mostly rely on manual experience and judgment, which are highly subjective and inefficient. With the development of computer technology and machine learning, some algorithm-based automatic anomaly recognition methods have gradually emerged. These methods extract the characteristic parameters of the chromatogram and compare and analyze them with the standard spectrum to achieve automatic anomaly recognition. However, existing methods often focus on the evaluation of overall similarity and ignore minor changes in the baseline part, resulting in insufficient recognition of anomalies such as baseline drift.

[0004] At present, the existing implementation schemes that are most similar to the present invention mainly focus on the chromatogram anomaly recognition method based on similarity calculation. These methods usually use indicators such as cosine similarity and Pearson correlation coefficient to calculate the similarity between the spectrum to be tested and the standard spectrum, and determine whether there is an anomaly by setting a threshold.

[0005] However, these methods have the following shortcomings: first, the similarity calculation is easily affected by the overall trend and it is difficult to accurately capture the slight changes in the baseline part; second, there is a lack of targeted detection methods for specific anomaly types such as baseline drift; third, the recognition results are greatly affected by the threshold setting and have poor robustness. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a transformer spectrum baseline anomaly identification method based on similarity and Wasserstein distance, which realizes automatic identification of anomalies such as baseline drift, ghost peaks or miscellaneous peaks in the spectrum by accurately calculating the similarity between the transformer oil chromatogram and the standard spectrum and the Wasserstein distance of the baseline part. The method combines the advantages of global similarity evaluation and local baseline difference detection, improves the accuracy and robustness of anomaly identification, and provides an efficient and reliable technical means for transformer fault diagnosis.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is: A transformer spectrum baseline anomaly identification method based on similarity and Wasserstein distance, step 1, using a filter to remove high-frequency noise of the chromatogram, removing baseline drift through a baseline correction algorithm, and performing operations including data cleaning, data integration, data transformation and data reduction to ensure data comparability and quality; Step 2: Select candidate reference atlases from the data source, perform preprocessing including denoising and smoothing, screen and verify the atlases, organize them into a standard atlas library, and maintain and update them; Step 3: Use the Pearson correlation coefficient method to calculate the overall similarity between the spectrum to be tested and the standard spectrum, and set a threshold to judge abnormalities; Step 4: Calculate the Wasserstein distance between the baseline of the spectrum to be tested and the standard baseline template, and set a threshold to determine baseline abnormality; Step 5: Combine the similarity and Wasserstein distance results to determine whether the spectrum baseline is abnormal; Step 6: Identify and output abnormal results, including similarity exceeding threshold, Wasserstein distance exceeding threshold and abnormal type. Preferably, in step 1, the data cleaning includes: (1) Filling missing values: using the prediction model to fill in missing values.

[0008] (2) Remove high-frequency noise: Use filters to remove high-frequency noise.

[0009] (3) Handling outliers: Use threshold setting or statistical distribution-based methods to make judgments and handle outliers.

[0010] (4) Ensure data accuracy and consistency: Ensure by correcting errors and inconsistencies.

[0011] The data integration is as follows: utilizing the data integration process to merge data from multiple data sources into a unified data set, while solving data redundancy and conflict issues to ensure data integrity and consistency.

[0012] The data transformation includes: (1) Scaling data: Scale the data to a specific range using proportional scaling.

[0013] (2) Adjust the distribution: Use the method of adjusting the data distribution to adjust the data distribution to a standard normal distribution, so that the mean is 0 and the standard deviation is 1.

[0014] (3) Convert data: Convert continuous data into discrete categories to facilitate data analysis or mining algorithms.

[0015] The data reduction is to reduce the amount of data by using dimensionality reduction, aggregation and other methods to reduce the amount of data while maintaining the original information of the data, thereby improving the efficiency of data processing and analysis.

[0016] Preferably, the sub-steps of step 2 are: Sub-step 2.1: Select representative atlases from existing databases as candidate reference atlases, ensuring that their sources are reliable and have been rigorously reviewed and verified; Sub-step 2.2: Perform denoising and smoothing preprocessing operations on the candidate reference atlas to improve the clarity and accuracy of the atlas; Sub-step 2.3: Standardize the spectra to ensure that different spectra are compared on the same scale and eliminate differences caused by instrumental and operational factors; Sub-step 2.4: Separate the baseline part and the peak part to facilitate subsequent analysis and comparison; Sub-step 2.5: Screen the pre-processed spectra according to the standards and exclude abnormal or non-compliant spectra; Sub-step 2.6: Verify the selected spectra by comparing them with the spectra of known samples and repeating the measurement method to ensure their representativeness and accuracy; Sub-step 2.7: Organize the screened and verified spectra into a standard spectra library, and number, classify and store them; Sub-step 2.8: Ensure the updating and maintenance of the standard spectrum library, add new standard spectrum in a timely manner and delete outdated or no longer used spectrum.

[0017] Preferably, the sub-steps of step 3 are: Sub-step 3.1: Calculate the overall similarity between the test spectrum and the standard spectrum using the Pearson correlation coefficient method; Sub-step 3.2: Setting similarity threshold; Sub-step 3.3: If the calculated similarity is lower than the threshold, it is considered as a possible anomaly.

[0018] Preferably, the method of step 3.3 is as follows: Use the Pearson correlation coefficient method to calculate the overall similarity between the spectrum to be tested and the standard spectrum; set a similarity threshold, and if it is lower than the threshold, it is considered as a possible abnormality; The Pearson correlation coefficient is an indicator that measures the degree of linear correlation between two variables X and Y. The value range of the Pearson correlation coefficient r is between -1 and 1, where: When r = 1, it means that X and Y are completely positively correlated, that is, one variable increases linearly as the other variable increases, or decreases linearly as the other variable decreases; When r=-1, it means that X and Y are completely negatively correlated, that is, one variable decreases linearly as the other variable increases, or increases linearly as the other variable decreases; When r = 0, it means there is no linear correlation between X and Y; The calculation formula of Pearson correlation coefficient is: ; in, and are the i-th observation values ​​of variables X and Y, respectively. and are the means of X and Y respectively, and n is the number of observations.

[0019] Preferably, the sub-steps of step 4 are: Sub-step 4.1: For the baseline part, obtain the baseline data of the spectrum to be tested and the data of the standard baseline template. These data are one-dimensional arrays, representing the baseline intensity at different wavelengths or frequencies. Sub-step 4.2: Based on the understanding of the tolerance of baseline drift and the statistical information obtained from a large amount of experimental data, set the Wasserstein distance threshold. Sub-step 4.3: Calculate the Wasserstein distance between the baseline of the spectrum to be tested and the standard baseline template: In Python, the scipy.stats.wasserstein_distance function in the scipy library can be used to directly calculate the Wasserstein distance between two one-dimensional arrays. Sub-step 4.4: If the calculated Wasserstein distance exceeds the threshold, it is considered as a baseline anomaly. At the same time, the Wasserstein distance can capture small differences between distributions and is suitable for detecting baseline drift.

[0020] Preferably, the calculation method of sub-step 4.3 is: For one-dimensional probability distributions P and Q, defined on the real number set R, their p-order Wasserstein distance is defined as: (P, Q) = ; Among them, Π(P,Q) is the set of all possible joint distributions of P and Q, and the marginal distributions of the joint distribution are P and Q respectively; represents the distance from x to y to the power of p; is the probability density of the joint distribution y at the point (x, y); p ≥ 1 is a real number, indicating that the p-order distance is considered; According to the actual application, two sets of data are obtained: one set is the baseline data of the spectrum to be tested, and the other set is the data of the standard baseline template; these data are one-dimensional arrays, representing the baseline intensity at different wavelengths or frequencies; in Python, the scipy.stats.wasserstein_distance function in the scipy library can be directly used to calculate the Wasserstein distance between two one-dimensional arrays.

[0021] Preferably, in sub-step 4.4, a threshold is set and whether the baseline is abnormal is determined. A Wasserstein distance threshold needs to be set according to the specific requirements of the application; this threshold should be based on the understanding of the tolerance of baseline drift and the statistical information obtained through a large amount of experimental data.

[0022] Preferably, in step 5, a comprehensive judgment is made based on the results of similarity and Wasserstein distance; if both indicate abnormality, it is confirmed that the spectrum has a baseline abnormality; if the similarity is lower than its threshold and the Wasserstein distance is higher than its threshold, it indicates that the spectrum has a baseline abnormality; if only one of the indicators indicates an abnormality and the other is within the normal range, further analysis or confirmation is required; if both are within the normal range, the baseline of the spectrum is considered to be normal.

[0023] Preferably, in step 6, the abnormality types include baseline drift, ghost peaks, and miscellaneous peaks; by developing a graphical user interface, it is convenient for users to upload chromatogram files, view abnormality identification results and perform interactive operations; an abnormality alarm function is provided to promptly notify users when a baseline abnormality is detected.

[0024] The present invention can achieve the following beneficial effects: 1. By combining the Pearson correlation coefficient and Wasserstein distance, the present invention can comprehensively evaluate the similarity between the spectrum to be tested and the standard spectrum from both the overall and local dimensions. The Pearson correlation coefficient is used to evaluate the similarity of the overall spectrum, while the Wasserstein distance pays special attention to the slight changes in the baseline part. This dual detection mechanism significantly improves the accuracy and reliability of anomaly detection.

[0025] 2. Wasserstein distance can capture subtle changes in the baseline part, and is particularly suitable for detecting minor anomalies such as baseline drift. By setting a suitable threshold, the present invention can promptly detect baseline anomalies such as baseline drift, ghost peaks or miscellaneous peaks, thereby avoiding misjudgment or missed detection caused by baseline problems.

[0026] 3. The present invention uses a variety of data preprocessing techniques, such as Gaussian filter to remove high-frequency noise, baseline correction algorithm (such as iterative polynomial fitting, wavelet transform, etc.) to remove baseline drift, as well as data cleaning, data integration, data transformation and data reduction. These preprocessing measures effectively improve the quality of data and ensure the accuracy of subsequent analysis.

[0027] 4. The present invention improves the robustness of the algorithm by adaptively training different types of chromatograms and abnormal situations. The threshold setting is optimized through a large amount of experimental data, and the algorithm performance is evaluated using indicators such as accuracy, recall rate, and F1 score, ensuring the stability and reliability of the system in various complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0029] The preferred solution is Figure 1 As shown in the figure, a transformer spectrum baseline anomaly identification method based on similarity and Wasserstein distance is used to identify baseline anomalies in the spectrum, such as baseline drift, ghost peaks or miscellaneous peaks, by calculating the similarity between the transformer oil chromatogram and the standard spectrum and the Wasserstein distance of the baseline part. Similarity is used to evaluate the similarity of the overall spectrum, while Wasserstein distance pays special attention to the slight changes in the baseline part. The combination of the two can detect anomalies more comprehensively. The specific method is as follows: 1. Data preprocessing: Using filters (such as Gaussian filters) can effectively remove high-frequency noise while retaining important chromatographic peak information. This is achieved by subtracting the mean and dividing by the standard deviation to ensure that the data is comparable between different chromatograms. Baseline correction algorithms (such as iterative polynomial fitting, wavelet transform, etc.) are used to remove baseline drift and identify and separate peak parts to facilitate subsequent analysis.

[0030] Data preprocessing mainly includes data cleaning, data integration, data transformation and data reduction.

[0031] Data cleaning: fill missing values ​​and use predictive models to fill them. Remove high-frequency noise through filtering technology. Judge and handle outliers in the data by setting thresholds or based on statistical distribution methods. Correct errors and inconsistencies in the data to ensure data accuracy and consistency.

[0032] Data integration is the process of combining data from multiple data sources into a unified data set. During the data integration process, it is necessary to solve problems such as data redundancy and data conflicts to ensure data integrity and consistency.

[0033] Data transformation: Scale the data so that it falls into a small specific interval, such as [0, 1] or [-1, 1]. Adjust the distribution of data to a standard normal distribution, with a mean of 0 and a standard deviation of 1. Convert continuous data into discrete categories to facilitate certain specific data analysis or data mining algorithms.

[0034] Data reduction is to reduce the amount of data through dimensionality reduction, aggregation and other methods while maintaining the original information of the data, so as to improve the efficiency of data processing and analysis.

[0035] 2. Establishment of standard spectrum: Select a series of standard chromatograms without abnormalities as references. Preprocess the standard spectra and extract the baseline template.

[0036] Select a series of representative spectra from the existing database as candidate reference spectra. Ensure that these spectra are derived from reliable data sources and have been strictly reviewed and verified. Perform preprocessing operations such as denoising and smoothing on the candidate reference spectra to improve the clarity and accuracy of the spectra. Standardize to ensure that different spectra are compared on the same scale and eliminate differences caused by instruments, operations, etc. Separate the baseline part and the peak part for subsequent analysis and comparison. Screen the preprocessed spectra according to the standards to exclude abnormal or non-compliant spectra. Verify the screened spectra to ensure their representativeness and accuracy. This can be done by comparing with the spectra of known samples, repeated measurements, etc. Organize the screened and verified spectra into a standard spectra library, and number, classify and store them. Ensure the update and maintenance of the standard spectra library, add new standard spectra in a timely manner, and delete outdated or no longer used spectra.

[0037] 3. Similarity calculation: The Pearson correlation coefficient method is used to calculate the overall similarity between the spectrum to be tested and the standard spectrum. A similarity threshold is set, and a value below the threshold is considered a possible anomaly.

[0038] The Pearson correlation coefficient is an indicator that measures the degree of linear correlation between two variables X and Y. This coefficient was proposed by Karl Pearson in the 1880s and is widely used in statistics. The value range of the Pearson correlation coefficient r is between -1 and 1, where: When r = 1, it means that X and Y are completely positively correlated, that is, one variable increases linearly as the other variable increases, or decreases linearly as the other variable decreases.

[0039] When r=-1, it means that X and Y are completely negatively correlated, that is, one variable decreases linearly as the other variable increases, or increases linearly as the other variable decreases.

[0040] When r = 0, it means there is no linear correlation between X and Y, but it does not mean there is no relationship between them. There may be other forms of relationship (such as nonlinear relationship).

[0041] The calculation formula of Pearson correlation coefficient is:

[0042] in, and are the i-th observation values ​​of variables X and Y, respectively. and are the means of X and Y respectively, and n is the number of observations.

[0043] 4. Wasserstein distance calculation: For the baseline part, the Wasserstein distance between the baseline of the spectrum to be tested and the standard baseline template is calculated. Wasserstein distance can capture the slight difference between distributions and is suitable for detecting baseline drift. Set the Wasserstein distance threshold, and if it exceeds the threshold, it is considered as baseline abnormality.

[0044] For one-dimensional probability distributions P and Q, defined on the real number set R, their p-order Wasserstein distance is defined as: (P, Q) = ; Among them, Π(P,Q) is the set of all possible joint distributions (or transmission plans) of P and Q, and the marginal distributions of these joint distributions are P and Q respectively. Represents the distance from x to y raised to the power of p. is the probability density of the joint distribution y at the point (x,y). p ≥ 1 is a real number, indicating that we are considering the p-order distance.

[0045] First, two sets of data are obtained according to the actual application: one set is the baseline data of the spectrum to be measured, and the other set is the data of the standard baseline template. These data are usually one-dimensional arrays, representing the baseline intensity at different wavelengths or frequencies. The calculation of Wasserstein distance can be simplified using some ready-made libraries. In Python, the scipy.stats.wasserstein_distance function in the scipy library can be used directly to calculate the Wasserstein distance between two one-dimensional arrays.

[0046] To set a threshold and determine whether the baseline is abnormal, you need to set a Wasserstein distance threshold based on the specific needs of the application. This threshold should be based on the understanding of the tolerance of baseline drift and the statistical information obtained from a large amount of experimental data.

[0047] V. Comprehensive judgment: Combine the results of similarity and Wasserstein distance for comprehensive judgment. If both indicate abnormality, it is confirmed that the spectrum has baseline abnormality. If the similarity is lower than its threshold and the Wasserstein distance is higher than its threshold, it indicates that the spectrum has baseline abnormality. If only one of the indicators indicates abnormality and the other is within the normal range, further analysis or confirmation is required, and there may be special circumstances or noise influence. If both are within the normal range, it can be considered that the baseline of the spectrum is normal.

[0048] 6. Output of abnormal identification results: Output abnormal identification results, including similarity, Wasserstein distance value and abnormal type (such as baseline drift, ghost peak, and miscellaneous peak). Develop a graphical user interface (GUI) to facilitate users to upload chromatogram files, view abnormal identification results and perform interactive operations. Provide abnormal alarm function to notify users in time when baseline abnormality is detected.

[0049] In this embodiment, the similarity is used to evaluate the overall consistency between the spectrum to be tested and the standard spectrum, and quickly screen out possible abnormal spectra. The Wasserstein distance is particularly targeted at the baseline part, capturing small distribution changes and accurately identifying abnormalities such as baseline drift.

[0050] In addition, the present invention adopts programming languages ​​such as Python and uses scientific computing libraries such as NumPy and SciPy to realize data processing and calculation. The similarity calculation method and Wasserstein distance calculation tool in the machine learning library are used. Deep learning technology, such as convolutional neural network (CNN), is introduced to automatically extract features and calculate similarity and distance. Unsupervised learning methods are used to automatically cluster standard spectra to improve the accuracy of similarity calculation. Adaptive training for different types of chromatograms and abnormal situations is increased to improve the robustness of the algorithm. Advanced methods in the field of anomaly detection, such as local outlier factor (LOF), are introduced and used in combination with existing methods.

[0051] Use known abnormal and normal chromatogram datasets for training and testing. Use indicators such as accuracy, recall, and F1 score to evaluate algorithm performance. Compare the recognition effects under different threshold settings and select the optimal threshold.

[0052] The above embodiments are only preferred technical solutions of the present invention and should not be regarded as limiting the present invention. The protection scope of the present invention shall be the technical solutions recorded in the claims, including equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A transformer spectrum baseline anomaly recognition method based on similarity and Wasserstein distance, characterized in that The following steps are involved: Step 1: Use a filter to remove high-frequency noise from the chromatogram, remove baseline drift through a baseline correction algorithm, and perform operations including data cleaning, data integration, data transformation, and data reduction to ensure data comparability and quality; Step 2: Select candidate reference atlases from the data source, perform preprocessing including denoising and smoothing, screen and verify the atlases, organize them into a standard atlas library, and maintain and update them; Step 3: Use the Pearson correlation coefficient method to calculate the overall similarity between the spectrum to be tested and the standard spectrum, and set a threshold to judge abnormalities; Step 4: Calculate the Wasserstein distance between the baseline of the spectrum to be tested and the standard baseline template, and set a threshold to determine baseline abnormality; Step 5: Combine the similarity and Wasserstein distance results to determine whether the spectrum baseline is abnormal; Step 6: Identify and output abnormal results, including similarity exceeding threshold, Wasserstein distance exceeding threshold and abnormal type.

2. According to claim 1, a transformer spectrum baseline anomaly identification method based on similarity and Wasserstein distance is characterized in that: In step 1, the data cleaning includes: (1) Filling missing values: using the prediction model to fill in missing values; (2) Remove high-frequency noise: Use filters to remove high-frequency noise; (3) Handling outliers: using threshold setting or statistical distribution-based methods to make judgments and handle them; (4) Ensure data accuracy and consistency: Ensure by correcting errors and inconsistencies; The data integration is as follows: using the data integration process to merge data from multiple data sources into a unified data set, while solving data redundancy and conflict problems to ensure data integrity and consistency; The data transformation includes: (1) Scaling data: scaling the data to a specific range using proportional scaling; (2) Adjust the distribution: Use the method of adjusting the data distribution to adjust the data distribution to a standard normal distribution, so that the mean is 0 and the standard deviation is 1; (3) Transforming data: Converting continuous data into discrete categories facilitates data analysis or mining algorithms; The data reduction is to reduce the amount of data, reduce the amount of data while maintaining the original information of the data using dimensionality reduction and aggregation methods, and improve the efficiency of data processing and analysis.

3. The method for identifying transformer spectrum baseline anomaly based on similarity and Wasserstein distance according to claim 1, characterized in that: The sub-steps of step 2 are: Sub-step 2.1: Select representative atlases from existing databases as candidate reference atlases, ensuring that their sources are reliable and have been rigorously reviewed and verified; Sub-step 2.2: Perform denoising and smoothing preprocessing operations on the candidate reference atlas to improve the clarity and accuracy of the atlas; Sub-step 2.3: Standardize the spectra to ensure that different spectra are compared on the same scale and eliminate differences caused by instrumental and operational factors; Sub-step 2.4: Separate the baseline part and the peak part to facilitate subsequent analysis and comparison; Sub-step 2.5: Screen the pre-processed spectra according to the standards and exclude abnormal or non-compliant spectra; Sub-step 2.6: Verify the selected spectra by comparing them with the spectra of known samples and repeating the measurement method to ensure their representativeness and accuracy; Sub-step 2.7: Organize the screened and verified spectra into a standard spectra library, and number, classify and store them; Sub-step 2.8: Ensure the updating and maintenance of the standard spectrum library, add new standard spectrum in a timely manner and delete outdated or no longer used spectrum.

4. The method for identifying transformer spectrum baseline anomaly based on similarity and Wasserstein distance according to claim 1, characterized in that: The sub-steps of step 3 are: Sub-step 3.1: Calculate the overall similarity between the test spectrum and the standard spectrum using the Pearson correlation coefficient method; Sub-step 3.2: Setting similarity threshold; Sub-step 3.3: If the calculated similarity is lower than the threshold, it is considered as a possible anomaly.

5. The method for identifying transformer spectrum baseline anomaly based on similarity and Wasserstein distance according to claim 1, characterized in that: Step 3.3: Use the Pearson correlation coefficient method to calculate the overall similarity between the spectrum to be tested and the standard spectrum; set a similarity threshold, and if it is lower than the threshold, it is considered as a possible abnormality; The Pearson correlation coefficient is an indicator that measures the degree of linear correlation between two variables X and Y. The value range of the Pearson correlation coefficient r is between -1 and 1, where: When r = 1, it means that X and Y are completely positively correlated, that is, one variable increases linearly as the other variable increases, or decreases linearly as the other variable decreases; When r=-1, it means that X and Y are completely negatively correlated, that is, one variable decreases linearly as the other variable increases, or increases linearly as the other variable decreases; When r = 0, it means there is no linear correlation between X and Y; The calculation formula of Pearson correlation coefficient is: ； in, and are the i-th observation values ​​of variables X and Y, respectively. and are the means of X and Y respectively, and n is the number of observations.

6. The method for identifying transformer spectrum baseline anomalies based on similarity and Wasserstein distance according to claim 1, characterized in that: The sub-steps of step 4 are: Sub-step 4.1: For the baseline part, obtain the baseline data of the spectrum to be tested and the data of the standard baseline template. These data are one-dimensional arrays, representing the baseline intensity at different wavelengths or frequencies; Sub-step 4.2: Based on the understanding of the tolerance of baseline drift and the statistical information obtained from a large amount of experimental data, set the Wasserstein distance threshold; Sub-step 4.3: Calculate the Wasserstein distance between the baseline of the spectrum to be tested and the standard baseline template: In Python, the scipy.stats.wasserstein_distance function in the scipy library can be used to directly calculate the Wasserstein distance between two one-dimensional arrays; Sub-step 4.4: If the calculated Wasserstein distance exceeds the threshold, it is considered as a baseline anomaly; at the same time, the Wasserstein distance can capture small differences between distributions and is suitable for detecting baseline drift.

7. The method for identifying transformer spectrum baseline anomaly based on similarity and Wasserstein distance according to claim 6, characterized in that: The calculation method of sub-step 4.3 is: For one-dimensional probability distributions P and Q, defined on the real number set R, their p-order Wasserstein distance is defined as: (P、Q)= ; Among them, Π(P,Q) is the set of all possible joint distributions of P and Q, and the marginal distributions of the joint distribution are P and Q respectively; represents the distance from x to y to the power of p; is the probability density of the joint distribution y at the point (x, y); p ≥ 1 is a real number, indicating that the p-order distance is considered; According to the actual application, two sets of data are obtained: one set is the baseline data of the spectrum to be tested, and the other set is the data of the standard baseline template; the two sets of data are one-dimensional arrays, representing the baseline intensity at different wavelengths or frequencies; in Python, the scipy.stats.wasserstein_distance function in the scipy library is used to directly calculate the Wasserstein distance between two one-dimensional arrays.

8. The method for identifying transformer spectrum baseline anomalies based on similarity and Wasserstein distance according to claim 6, characterized in that: In sub-step 4.4, a threshold is set to determine whether the baseline is abnormal. A Wasserstein distance threshold needs to be set according to the specific needs of the application. This threshold is based on the understanding of the tolerance of baseline drift and the statistical information obtained from a large amount of experimental data.

9. The method for identifying transformer spectrum baseline anomaly based on similarity and Wasserstein distance according to claim 1, characterized in that: In step 5, a comprehensive judgment is made based on the results of similarity and Wasserstein distance; if both indicate abnormality, it is confirmed that the spectrum has a baseline abnormality; if the similarity is lower than its threshold and the Wasserstein distance is higher than its threshold, it indicates that the spectrum has a baseline abnormality; if only one of the indicators indicates an abnormality and the other is within the normal range, further analysis or confirmation is required; if both are within the normal range, the baseline of the spectrum is considered to be normal.

10. The method for identifying transformer spectrum baseline anomalies based on similarity and Wasserstein distance according to claim 1, characterized in that: In step 6, the abnormality types include baseline drift, ghost peaks, and miscellaneous peaks; by developing a graphical user interface, it is convenient for users to upload chromatogram files, view abnormality identification results, and perform interactive operations; an abnormality alarm function is provided to notify users in a timely manner when a baseline abnormality is detected.

Images