Method for identifying gas components in transformer oil and monitoring concentration of gas components in transformer oil based on spectral analysis

Abstract

The invention discloses a method for identifying gas components in transformer oil and monitoring concentration based on spectral analysis, relates to the technical field of transformer oil, and aims to solve the problem of inaccurate detection of abnormal conditions of gas components and gas concentration of transformer oil. According to the method, whether the detected concentration value is in a normal range, a critical range or an abnormal range or not can be intuitively judged through curve overlapping comparison results, the peak area is calculated by selecting a triangle method or a numerical integration method according to the symmetry of peak shapes, and the calculation accuracy and applicability are ensured through flexible selection; spectral data without target gas can be obtained through background scanning, and a comparison reference is provided for subsequent sample scanning. Sample scanning is directly carried out on target detection gas, and spectral characteristics of the target detection gas can be accurately obtained. The method of multiple scanning and averaging effectively reduces random errors, and improves the stability and repeatability of data.

Description

Technical Field

[0001] The invention relates to the technical field of transformer oil, in particular to a method for identifying gas components in transformer oil and monitoring their concentration based on spectral analysis. Background Art

[0002] Identification of gas components in transformer oil refers to the detection and identification of various gas components and their contents dissolved in transformer oil through specific analysis methods and techniques.

[0003] The Chinese patent document with publication number CN110132864A discloses a method for detecting gas in transformer oil, which mainly transmits a laser light source into a gas chamber through an optical fiber, absorbs the detection gas in the gas chamber, and the light intensity is attenuated. The light absorbed by the gas chamber is transmitted to a light detector through an optical fiber, and photoelectric conversion is performed by the light detector. The signal after photoelectric conversion is input into a signal processing circuit module for signal processing, and the gas composition and concentration in the gas chamber are analyzed. Although the above patent document solves the problem of gas composition detection, the following problems still exist in actual operation: 1. Failure to adopt a more complete gas collection method and to further process the collected gas resulted in inaccurate gas collection in transformer oil.

[0004] 2. There is no targeted analysis of the gas, which makes it impossible to directly obtain the gas components.

[0005] 3. The concentration of gas components in transformer oil is not monitored, so targeted warnings cannot be issued based on the gas concentration. Summary of the invention

[0006] The purpose of the present invention is to provide a method for identifying gas components and monitoring concentration in transformer oil based on spectral analysis. Through the curve overlap comparison results, it can be intuitively judged whether the detection concentration value is in the normal range, critical range or abnormal range. According to the symmetry of the peak shape, the triangle method or the numerical integration method is selected to calculate the peak area. This flexible selection ensures the accuracy and applicability of the calculation. The background scan can obtain the spectral data when there is no target gas, providing a comparison benchmark for subsequent sample scanning. The sample scan is directly performed on the target detection gas, and its spectral characteristics can be accurately obtained. The method of multiple scanning and averaging effectively reduces random errors, improves the stability and repeatability of the data, and can solve the problems in the prior art.

[0007] To achieve the above object, the present invention provides the following technical solutions: The method for identifying gas components and monitoring concentration in transformer oil based on spectral analysis includes the following steps: S1: Sample processing and collection: inject the oil sample in the transformer into the degassing device for degassing, collect the gas after degassing, and perform equipment inspection on the degassing device before degassing, and mark the collected gas as target detection gas data; S2: Spectral data acquisition and processing: The target detected gas data is collected by using an infrared spectrometer to collect gas molecule data, and the collected gas molecule data is preprocessed to obtain gas spectrum data; S3: Spectral data feature analysis: extracting feature data from gas spectral data, performing feature screening on the extracted feature data, generating a gas feature vector after feature screening, and marking the generated gas feature vector as gas feature analysis data; When extracting characteristic data from the gas spectrum data in S3, the extracted characteristic data include peak area, peak height, peak width and peak shape.

[0008] Preferably, in S1, the oil sample in the transformer is injected into a degassing device for degassing, the gas after degassing is collected, and the degassing device is tested before degassing, including: The degassing device is a vacuum degassing device. Before degassing the oil sample, the degassing device shall be inspected. The equipment inspection includes appearance inspection, electrical inspection, vacuum system inspection, heating inspection, filtration inspection and safety inspection. After the equipment is tested and qualified, use the sampling equipment to take the oil sample from the transformer and inject the taken oil sample into the sample chamber of the degassing device; After the oil sample is injected, the working parameters of the degassing device are set. The working parameters include working time and temperature. The temperature is between 40°C and 60°C, and the time is 30 minutes to 1 hour. The gas released during the degassing process is transferred to the gas collection bottle through a conduit; The gas in the gas collection bottle is marked as the target detection gas data.

[0009] Preferably, in S2, the target detected gas data is collected using an infrared spectrometer for gas molecule data collection, and the collected gas molecule data is preprocessed, including: Before the infrared spectrometer collects gas molecule data, it is first calibrated, including optical system calibration, background calibration and wavelength calibration; injecting the collected target detection gas data into the sample chamber of the infrared spectrometer; After the injection is completed, the infrared spectrometer is started to scan the target detection gas data in the sample chamber, and the scanning includes background scanning, sample scanning and multiple scanning; Background scanning is to record the spectral data when there is no target detection gas data; sample scanning is to scan the target detection gas data; multiple scanning is to scan the target detection gas data multiple times and take the average value; After the target detection gas data in the sample chamber is scanned, the gas molecule data of the target detection gas data is obtained; The gas molecule data are preprocessed, including baseline correction, spectrum normalization, denoising and data smoothing; After data preprocessing, gas spectrum data is obtained.

[0010] Preferably, in S3, feature data is extracted from the gas spectrum data, and feature screening is performed on the extracted feature data. After feature screening, a gas feature vector is generated, and the generated gas feature vector includes: When extracting characteristic data from gas spectrum data, the extracted characteristic data include peak area, peak height, peak width and peak shape; Peak shape is the analysis of the shape of the peak in the gas spectrum data; peak height is the measurement of the height of each absorption peak in the gas spectrum data; peak width is the analysis of the width of the peak in the gas spectrum data; The peak area is calculated based on the peak shape, including the triangle method and the numerical integration method. The triangle method is used for the calculation of symmetrical peaks, and the numerical integration method is used for the calculation of asymmetrical peaks. The calculation formula of the triangle method is as follows: ; The calculation formula of numerical integration method is as follows: ;

[0011] in, It is expressed as the width of each cell; Expressed as The function value of sampling points; It is expressed as the total number of sampling points; Perform feature screening on the extracted feature data, which includes correlation analysis and principal component analysis; The data after feature screening are combined into a feature vector, each feature vector represents a spectrum sample, and each spectrum sample is standardized to obtain a gas feature vector after the standardization process.

[0012] Preferably, it includes: S4: Gas identification and collation: Identify the gas type of the gas characteristic analysis data according to the gas identification model, and generate visualization data for the identified gas type. After the visualization data is generated, the gas type data is obtained; Divide the gas feature vectors into training set and test set; Retrieving a gas recognition model from a database, wherein the gas recognition model is a deep learning model; The gas recognition model is trained using the training set of gas feature vectors; The model training process is to input the feature vector in the training set into the gas recognition model, calculate the loss function after the feature vector is input, and use the k-fold cross-validation method to verify the gas recognition model after the loss function is calculated. After verification, the model training is completed; Use the test set to evaluate the performance of the trained gas recognition model. The performance evaluation includes accuracy evaluation, recall evaluation, F1 score evaluation, and confusion matrix evaluation. The gas feature vector is input into the gas identification model after performance evaluation to predict the type of gas, and the confidence of each gas type is obtained after the type prediction; Visualize the confidence of each gas type, including bar charts, scatter plots and boundary plots; Generate a result report using the visualized converted data, and annotate the generated result report as gas type data.

[0013] Preferably, it includes: S5: quantitative concentration analysis: retrieve gas concentration data from gas type data, and perform sample concentration analysis on the retrieved concentration data to obtain gas concentration data after concentration analysis; The gas type data includes the type name, concentration value, unit and detection time of each gas type. The concentration value is obtained by performing concentration value detection on the sample scan of the infrared spectrometer. Perform quantitative analysis based on the obtained concentration values; The quantitative analysis process includes: The standard concentration data of each gas type is retrieved from the database, and a standard curve is established for the standard concentration data; Establish a detection curve using the acquired concentration values; Overlapping and comparing the established standard curve with the established detection curve; According to the curve overlap comparison results, determine whether the test curve overlaps with the standard curve; If the detection curve is lower than the standard curve, the concentration value corresponding to the detection curve is within the normal range of the gas type; If the detection curve coincides with the standard curve, the concentration value corresponding to the detection curve is the critical range of the gas type; If the detection curve exceeds the standard curve, the concentration value corresponding to the detection curve is the abnormal range of the gas type; Finally, the curve overlap comparison results are marked as gas concentration data.

[0014] Preferably, it includes: S6: Gas concentration judgment and warning: formulate different concentration warning ranges for different gas types in the gas concentration data, and make warning responses to the concentration data of different gas types in the gas concentration data according to the formulated different concentration warning ranges; According to the curve overlap comparison results of the gas concentration data, the analysis results of each gas type and the concentration value corresponding to the gas type are obtained; The analysis results include normal concentration, critical concentration and abnormal concentration; Formulate an early warning range for each gas type, and the early warning range is formulated by setting critical thresholds and abnormal thresholds for each gas type; The critical concentration and abnormal concentration data in the analysis results are used for early warning response according to the established early warning range; The early warning response process includes: Confirm the gas type data in the critical concentration and abnormal concentration, and integrate the confirmed gas type data and the concentration value of the data; The integrated data is divided into warning levels according to the critical threshold and abnormal threshold in the established warning range; The warning levels include level one, level two and level three; Different intensities of early warning will be issued according to different warning levels.

[0015] Preferably, the method of performing an early warning response on the concentration data of different gas types in the gas concentration data according to the established different concentration early warning ranges also includes: Extract the warning response result of gas concentration, and extract the number of gas types corresponding to each warning level in the gas type data; Compare the number of gas types corresponding to each warning level with the corresponding preset number threshold; When the number of gas types in the first-level warning is lower than the corresponding number threshold, it is determined whether the number of gas types in the second-level warning and the third-level warning exceeds the corresponding number threshold; When the number of gas types in the third-level warning exceeds the corresponding number threshold, the detected gas data is marked as dangerous for gas concentration; When the number of gas types in the third-level warning does not exceed the corresponding number threshold, but the number of gas types in the second-level warning exceeds the corresponding number threshold, the gas concentration hazard judgment is performed by combining the gas concentration value corresponding to the gas type in the second-level warning with the gas concentration value corresponding to the gas type in the first-level warning; When the number of gas types in the third-level warning does not exceed the corresponding number threshold, and the number of gas types in the second-level warning does not exceed the corresponding number threshold, the gas concentration hazard judgment is made by using the gas concentration values ​​corresponding to the gas types in the third-level warning and the second-level warning combined with the gas concentration values ​​corresponding to the gas types in the first-level warning.

[0016] Preferably, the gas concentration value corresponding to the gas type of the second-level warning is combined with the gas concentration value corresponding to the gas type of the first-level warning to perform gas concentration hazard determination, including: When the number of gas types in the third-level warning does not exceed the corresponding number threshold, but the number of gas types in the second-level warning exceeds the corresponding number threshold, the gas concentration value corresponding to the gas type in the first-level warning is retrieved; According to the gas concentration value corresponding to the gas type of the first-level warning, the concentration difference between the gas concentration value corresponding to the gas type of the first-level warning and the critical threshold is obtained; Retrieving the critical threshold, and performing ratio processing on the concentration difference between the gas concentration value corresponding to the gas type of the first-level warning and the critical threshold and the critical threshold, respectively, to obtain the gas concentration ratio corresponding to the gas type of the first-level warning as the first gas concentration ratio data; Retrieve the gas concentration value corresponding to the gas type of the second-level warning; According to the gas concentration value corresponding to the gas type of the second-level warning, the concentration difference between the gas concentration value corresponding to the gas type of the second-level warning and the abnormal threshold is obtained; Retrieving the abnormal threshold, and performing ratio processing on the concentration difference between the gas concentration value corresponding to the gas type of the secondary warning and the abnormal threshold and the abnormal threshold, respectively, to obtain the gas concentration ratio corresponding to the gas type of the secondary warning as the second gas concentration ratio data; Obtaining a first gas concentration coefficient by combining the first gas concentration ratio data with a gas concentration standard deviation of the second gas concentration ratio data; Wherein, the first gas concentration coefficient is obtained by the following formula: ;

[0017] Among them, R 01 represents the first gas concentration coefficient; Y 01 represents the average value of the gas concentration ratio corresponding to the first gas concentration ratio data; Y 01max represents the maximum value of the gas concentration ratio corresponding to the first gas concentration ratio data; δ represents the standard deviation of the gas concentration corresponding to the first gas concentration ratio data; σ represents the standard deviation of the gas concentration corresponding to the second gas concentration ratio data; Obtaining a second gas concentration coefficient using the second gas concentration ratio data; Wherein, the second gas concentration coefficient is obtained by the following formula: ;

[0018] Among them, R 02 represents the second gas concentration coefficient; Y 02 represents the average value of the gas concentration ratio corresponding to the second gas concentration ratio data; Y 02max represents the maximum value of the gas concentration ratio corresponding to the second gas concentration ratio data; δ represents the standard deviation of the gas concentration corresponding to the first gas concentration ratio data; σ represents the standard deviation of the gas concentration corresponding to the second gas concentration ratio data; comparing the first gas concentration coefficient and the second gas concentration coefficient; When the second gas concentration coefficient is not lower than the first gas concentration coefficient, the detected gas data is determined to be dangerous, and a gas concentration danger mark is performed on the detected gas data.

[0019] Preferably, the gas concentration hazard determination is performed using the gas concentration values ​​corresponding to the gas types of the third-level warning and the second-level warning combined with the gas concentration values ​​corresponding to the gas types of the first-level warning, including: When the number of gas types in the third-level warning does not exceed the corresponding number threshold, and the number of gas types in the second-level warning does not exceed the corresponding number threshold, the difference between the gas concentration value corresponding to the gas type in the first-level warning and the abnormal threshold is processed with the critical threshold to obtain the ratio data between the difference between the gas concentration value corresponding to the gas type in the first-level warning and the critical threshold and the abnormal threshold as the third gas concentration ratio data; Obtaining a third gas concentration coefficient using the third gas concentration ratio data; The third gas concentration coefficient is obtained by the following formula: ;

[0020] Among them, R 03 Indicates the third gas concentration coefficient; Y 01 represents the average value of the gas concentration ratio corresponding to the first gas concentration ratio data; Y 01max represents the maximum value of the gas concentration ratio corresponding to the first gas concentration ratio data; Y 01min represents the minimum value of the gas concentration ratio corresponding to the first gas concentration ratio data; δ represents the standard deviation of the gas concentration corresponding to the first gas concentration ratio data; σ represents the standard deviation of the gas concentration corresponding to the second gas concentration ratio data; Performing ratio processing on the difference between the gas concentration value corresponding to the gas type of the secondary warning and the abnormal threshold and the abnormal threshold, obtaining the ratio data between the difference between the gas concentration value corresponding to the gas type of the secondary warning and the abnormal threshold and the abnormal threshold as the second gas concentration ratio data; Performing ratio processing on the difference between the gas concentration value corresponding to the gas type of the third-level warning and the abnormal threshold and the abnormal threshold, obtaining the ratio data between the difference between the gas concentration value corresponding to the gas type of the third-level warning and the abnormal threshold and the abnormal threshold as the third gas concentration ratio data; Obtaining a comprehensive gas concentration coefficient using the second gas concentration ratio data and the third gas concentration ratio data; The comprehensive gas concentration coefficient is obtained by the following formula: ;

[0021] Among them, R z Indicates the comprehensive gas concentration coefficient; Y 02 represents the average value of the gas concentration ratio corresponding to the second gas concentration ratio data; Y 02max represents the maximum value of the gas concentration ratio corresponding to the second gas concentration ratio data; δ represents the standard deviation of the gas concentration corresponding to the first gas concentration ratio data; σ represents the standard deviation of the gas concentration corresponding to the second gas concentration ratio data; Y 03 represents the average value of the gas concentration ratio corresponding to the third gas concentration ratio data; Y 03max represents the maximum value of the gas concentration ratio corresponding to the third gas concentration ratio data; μ represents the standard deviation of the gas concentration corresponding to the third gas concentration ratio data; comparing the third gas concentration coefficient with the comprehensive gas concentration coefficient; When the comprehensive gas concentration coefficient is not lower than the third gas concentration coefficient, the detected gas data is determined to be dangerous, and a gas concentration danger mark is performed on the detected gas data.

[0022] Compared with the prior art, the present invention has the following beneficial effects: 1. The method for identifying gas components and monitoring concentration in transformer oil based on spectral analysis provided by the present invention, the gas released during the degassing process is transferred to the gas collection bottle through a conduit, which realizes the effective collection and preservation of the gas, provides a reliable sample for subsequent gas analysis, and the background scan can obtain the spectral data when there is no target gas, providing a comparison benchmark for subsequent sample scanning. The sample scan is directly performed on the target detection gas, and its spectral characteristics can be accurately obtained. The method of multiple scanning and averaging effectively reduces random errors and improves the stability and repeatability of the data.

[0023] 2. The method for identifying gas components and monitoring concentration in transformer oil based on spectral analysis provided by the present invention uses a deep learning model as a gas identification model, which can process complex data features and capture subtle differences in gas feature vectors, thereby improving the accuracy and robustness of identification. The triangle method is suitable for symmetrical peaks, and the numerical integration method can better process asymmetrical peaks, thereby improving the accuracy of peak area calculation. The triangle method or the numerical integration method is selected to calculate the peak area according to the symmetry of the peak shape. This flexible choice ensures the accuracy and applicability of the calculation.

[0024] 3. The method for identifying gas components and monitoring concentration in transformer oil based on spectral analysis provided by the present invention is automated through programming, reducing errors caused by human operation. Through the curve overlap comparison results, it can be intuitively judged whether the detected concentration value is in the normal range, critical range or abnormal range. The warning level is subdivided into first-level warning, second-level warning and third-level warning. Different levels of warnings correspond to different intensities and urgency, which helps managers take corresponding response measures according to the warning level. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a schematic diagram of the steps of monitoring the concentration of gas components in transformer oil according to the present invention. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] In order to solve the problem that in the prior art, when collecting gas from transformer oil samples, a more perfect gas collection method is not adopted and the collected gas is not further processed, resulting in inaccurate collection of gas in transformer oil, please refer to Figure 1 , this embodiment provides the following technical solutions: The method for identifying gas components and monitoring concentration in transformer oil based on spectral analysis includes the following steps: S1: Sample processing and collection: inject the oil sample in the transformer into the degassing device for degassing, collect the gas after degassing, and perform equipment inspection on the degassing device before degassing, and mark the collected gas as target detection gas data; Among them, the working parameters of the degassing device can be accurately set according to needs, and the setting of temperature range and time range can not only ensure the degassing effect, but also avoid damage to the sample due to improper parameters; S2: Spectral data acquisition and processing: The target detected gas data is collected by using an infrared spectrometer to collect gas molecule data, and the collected gas molecule data is preprocessed to obtain gas spectrum data; Among them, the method of multiple scanning and averaging effectively reduces random errors and improves the stability and repeatability of data; S3: Spectral data feature analysis: extracting feature data from gas spectral data, performing feature screening on the extracted feature data, generating a gas feature vector after feature screening, and marking the generated gas feature vector as gas feature analysis data; Among them, feature screening through correlation analysis and principal component analysis helps to remove redundant information and retain the most representative features; S4: Gas identification and collation: Identify the gas type of the gas characteristic analysis data according to the gas identification model, and generate visualization data for the identified gas type. After the visualization data is generated, the gas type data is obtained; Among them, the performance of the model is evaluated by calculating the loss function and verified by using the k-fold cross-validation method, which helps to ensure the generalization ability of the model; S5: quantitative concentration analysis: retrieve gas concentration data from gas type data, and perform sample concentration analysis on the retrieved concentration data to obtain gas concentration data after concentration analysis; Among them, by comparing the results of curve overlap, it is possible to intuitively determine whether the detected concentration value is within the normal range, critical range or abnormal range; S6: Gas concentration judgment and warning: formulate different concentration warning ranges for different gas types in the gas concentration data, and make warning responses to the concentration data of different gas types in the gas concentration data according to the formulated different concentration warning ranges; Among them, different levels of warnings correspond to different intensities and urgency, which helps managers take corresponding response measures based on the warning level.

[0028] In S1, the oil sample in the transformer is injected into the degassing device for degassing, the gas after degassing is collected, and the degassing device is tested before degassing, including: The degassing device is a vacuum degassing device. Before degassing the oil sample, the degassing device shall be inspected. The equipment inspection includes appearance inspection, electrical inspection, vacuum system inspection, heating inspection, filtration inspection and safety inspection. After the equipment is tested and qualified, use the sampling equipment to take the oil sample from the transformer and inject the taken oil sample into the sample chamber of the degassing device; After the oil sample is injected, the working parameters of the degassing device are set. The working parameters include working time and temperature. The temperature is between 40°C and 60°C, and the time is 30 minutes to 1 hour. The gas released during the degassing process is transferred to the gas collection bottle through a conduit; The gas in the gas collection bottle is marked as the target detection gas data.

[0029] Specifically, before degassing, a comprehensive equipment inspection is carried out on the degassing device, including inspections of appearance, electrical, vacuum system, heating, filtration and safety, etc., to ensure the stability and safety of the equipment and avoid experimental errors or safety accidents caused by equipment failure. A special sampling device is used to take oil samples from the transformer and accurately inject the oil samples into the sample chamber of the degassing device to ensure the representativeness and accuracy of the samples. The working parameters of the degassing device (such as working time and temperature) can be accurately set as needed, and the setting of the temperature range (40°C-60°C) and time range (30 minutes-1 hour) not only ensures the degassing effect, but also avoids damage to the sample due to improper parameters. The gas released during the degassing process is transferred to the gas collection bottle through a conduit, realizing the effective collection and preservation of the gas, and providing a reliable sample for subsequent gas analysis. It is not only suitable for the degassing of transformer oil samples, but can also be extended to other oil samples that require degassing, and has certain versatility and practicality.

[0030] In S2, the target gas detection data is collected by using an infrared spectrometer to collect gas molecule data, and the collected gas molecule data is preprocessed, including: Before the infrared spectrometer collects gas molecule data, it is first calibrated, including optical system calibration, background calibration and wavelength calibration; injecting the collected target detection gas data into the sample chamber of the infrared spectrometer; After the injection is completed, the infrared spectrometer is started to scan the target detection gas data in the sample chamber, and the scanning includes background scanning, sample scanning and multiple scanning; Background scanning is to record the spectral data when there is no target detection gas data; sample scanning is to scan the target detection gas data; multiple scanning is to scan the target detection gas data multiple times and take the average value; After the target detection gas data in the sample chamber is scanned, the gas molecule data of the target detection gas data is obtained; The gas molecule data are preprocessed, including baseline correction, spectrum normalization, denoising and data smoothing; After data preprocessing, gas spectrum data is obtained.

[0031] Specifically, optical system calibration ensures that the optical performance of the instrument is in the best state, providing a basis for accurate measurement. Background calibration helps to eliminate the influence of the instrument itself and the environmental background on the measurement results. Wavelength calibration ensures the accuracy of spectral data, making the analysis results more reliable. Background scanning can obtain spectral data when there is no target gas, providing a comparison benchmark for subsequent sample scanning. Sample scanning is directly performed on the target detection gas, and its spectral characteristics can be accurately obtained. The method of multiple scanning and averaging effectively reduces random errors and improves the stability and repeatability of the data. Baseline correction can eliminate baseline drift in spectral data, making spectral features more obvious. Spectral normalization processing makes the spectral data between different samples comparable, which is convenient for subsequent analysis. Denoising and data smoothing further reduce noise and fluctuations in the data, improve the signal-to-noise ratio and clarity of the data. Through comprehensive calibration and fine scanning process, as well as systematic data preprocessing, the gas spectral data finally obtained has high accuracy and reliability.

[0032] In order to solve the problem that in the prior art, after gas collection from transformer oil, no targeted analysis of the gas is performed, resulting in the inability to directly obtain the gas components, please refer to Figure 1 , this embodiment provides the following technical solutions: In S3, feature data is extracted from gas spectrum data, and feature screening is performed on the extracted feature data. After feature screening, a gas feature vector is generated. The generated gas feature vector includes: When extracting characteristic data from gas spectrum data, the extracted characteristic data include peak area, peak height, peak width and peak shape; Peak shape is to analyze the shape of the peak in the gas spectrum data; peak height is to measure the height of each absorption peak in the gas spectrum data; peak width is to analyze the width of the peak in the gas spectrum data; The peak area is calculated based on the peak shape, including the triangle method and the numerical integration method. The triangle method is used for the calculation of symmetrical peaks, and the numerical integration method is used for the calculation of asymmetrical peaks. The calculation formula of the triangle method is as follows: ; The calculation formula of numerical integration method is as follows: ;

[0033] in, It is expressed as the width of each cell; Expressed as The function value of sampling points; It is expressed as the total number of sampling points; Perform feature screening on the extracted feature data, which includes correlation analysis and principal component analysis; The data after feature screening are combined into a feature vector. Each feature vector represents a spectrum sample. Each spectrum sample is standardized to obtain a gas feature vector after standardization. .

[0034] Specifically, it covers multiple key features such as peak area, peak height, peak width and peak shape. These features can comprehensively reflect the main information of gas spectra and provide a rich data basis for subsequent analysis. According to the symmetry of the peak shape, the triangle method or the numerical integration method is selected to calculate the peak area. This flexible choice ensures the accuracy and applicability of the calculation. The triangle method is suitable for symmetrical peaks, and the numerical integration method can better handle asymmetrical peaks, thereby improving the accuracy of peak area calculation. Feature screening through correlation analysis and principal component analysis helps to remove redundant information and retain the most representative features. This can not only reduce the dimension of the data and reduce the computational complexity, but also improve the generalization ability and prediction accuracy of the model. Standardization of the data after feature screening can eliminate the dimensional differences between different features, making the data more unified and comparable. This helps the stable operation and performance improvement of the subsequent algorithm. The gas feature vector after feature extraction, screening and standardization can be used as the input of the machine learning algorithm. These high-quality data are helpful for the training and optimization of the algorithm, thereby improving the prediction performance and generalization ability of the model.

[0035] According to the gas identification model in S4, the gas type is identified for the gas characteristic analysis data, and the visual data of the identified gas type is generated, including: Divide the gas feature vectors into training set and test set; Retrieving a gas recognition model from a database, wherein the gas recognition model is a deep learning model; The gas recognition model is trained using the training set of gas feature vectors; The model training process is to input the feature vector in the training set into the gas recognition model, calculate the loss function after the feature vector is input, and use the k-fold cross-validation method to verify the gas recognition model after the loss function is calculated. After verification, the model training is completed; Use the test set to evaluate the performance of the trained gas recognition model. The performance evaluation includes accuracy evaluation, recall evaluation, F1 score evaluation, and confusion matrix evaluation. The gas feature vector is input into the gas identification model after performance evaluation to predict the type of gas, and the confidence of each gas type is obtained after the type prediction; Visualize the confidence of each gas type, including bar charts, scatter plots and boundary plots; Generate a result report using the visualized converted data, and annotate the generated result report as gas type data.

[0036] Specifically, the gas feature vectors are clearly divided into training sets and test sets to ensure the independence and accuracy of model training and evaluation. This data division method helps to avoid overfitting and makes the model evaluation results more credible. The deep learning model is used as the gas recognition model, which can handle complex data features and capture the subtle differences in the gas feature vectors, thereby improving the accuracy and robustness of recognition. During the model training process, the performance of the model is evaluated by loss function calculation and verified by k-fold cross validation, which helps to ensure the generalization ability of the model and reduce the deviation caused by improper data division. When evaluating the performance of the trained gas recognition model, multiple evaluation indicators such as accuracy, recall rate, F1 score and confusion matrix are used. These indicators can fully reflect the performance of the model and provide a basis for model optimization. The confidence of each gas type is predicted and visualized, which helps users understand the prediction results of the model more intuitively and improve the accuracy of decision-making. At the same time, the diversity of visualization transformations (bar charts, scatter plots and boundary maps) can meet the needs of different users.

[0037] In order to solve the problem that the existing technology does not monitor the concentration of gas components in transformer oil, and thus cannot provide targeted warnings based on gas concentration conditions, please refer to Figure 1 , this embodiment provides the following technical solutions: In S5, the concentration data of the gas is retrieved from the gas type data, and the retrieved concentration data is used for sample concentration analysis, including: The gas type data includes the type name, concentration value, unit and detection time of each gas type. The concentration value is obtained by performing concentration value detection on the sample scan of the infrared spectrometer. Perform quantitative analysis based on the obtained concentration values; The quantitative analysis process includes: The standard concentration data of each gas type is retrieved from the database, and a standard curve is established for the standard concentration data; Establish a detection curve using the acquired concentration values; Overlapping and comparing the established standard curve with the established detection curve; According to the curve overlap comparison results, determine whether the test curve overlaps with the standard curve; If the detection curve is lower than the standard curve, the concentration value corresponding to the detection curve is within the normal range of the gas type; If the detection curve coincides with the standard curve, the concentration value corresponding to the detection curve is the critical range of the gas type; If the detection curve exceeds the standard curve, the concentration value corresponding to the detection curve is the abnormal range of the gas type; Finally, the curve overlap comparison results are marked as gas concentration data.

[0038] Specifically, the concentration value is detected by infrared spectrometer to ensure the accuracy and scientificity of the data. The retrieval of standard concentration data and the establishment of standard curves provide a reliable benchmark for quantitative analysis. The steps of retrieving data from the database, establishing standard curves and detection curves, and curve overlap comparison realize the standardization of the process. The whole process can be automated through programming to reduce the errors caused by human operation. Through the curve overlap comparison results, it can be intuitively judged whether the detected concentration value is in the normal range, critical range or abnormal range. The final result is marked in the form of gas concentration data to facilitate subsequent data processing and analysis. Each step can be adjusted and optimized according to actual needs, such as improving the detection method, optimizing the curve establishment algorithm, etc. It can be integrated with other data analysis systems or platforms to achieve more advanced data processing and analysis functions. By providing accurate concentration analysis results, it provides a scientific basis for decision makers and helps to take necessary measures in time to deal with potential environmental or safety issues.

[0039] Different concentration warning ranges are formulated for different gas types in the gas concentration data in S2, and warning responses are made to the concentration data of different gas types in the gas concentration data according to the formulated different concentration warning ranges, including: According to the curve overlap comparison results of the gas concentration data, the analysis results of each gas type and the concentration value corresponding to the gas type are obtained; The analysis results include normal concentration, critical concentration and abnormal concentration; Formulate an early warning range for each gas type, and the early warning range is formulated by setting critical thresholds and abnormal thresholds for each gas type; The warning range parameter table is as follows: For example: ;

[0040] The critical concentration and abnormal concentration data in the analysis results are used for early warning response according to the established early warning range; The early warning response process includes: Confirm the gas type data in the critical concentration and abnormal concentration, and integrate the confirmed gas type data and the concentration value of the data; The integrated data is divided into warning levels according to the critical threshold and abnormal threshold in the established warning range; The warning levels include level one, level two and level three; Provide warning prompts of different intensities according to different warning levels; The early warning response process parameter table is as follows: For example: ;

[0041] The following are examples of early warning responses: ;

[0042] Specifically, different concentration warning ranges are formulated for different types of gases, fully considering the different dangers and concentration sensitivities that different gases may have, improving the pertinence and accuracy of the warning. By setting critical thresholds and abnormal thresholds, the concentration data is divided into three levels: normal, critical and abnormal, which helps managers to clearly understand the current safety status of gas concentrations. The warning response process includes multiple steps such as confirming the type of gas, integrating data, dividing the warning level and warning prompts. Each step has a clear operating guide to ensure that the warning response can be executed quickly and accurately. The warning level is subdivided into level one, level two and level three. Different levels of warnings correspond to different intensities and urgency, which helps managers take corresponding response measures according to the warning level. Through timely and accurate warnings, the solution can sound an alarm before the gas concentration reaches a dangerous level, thereby reminding relevant personnel to take timely measures to avoid safety accidents.

[0043] Specifically, according to the different concentration warning ranges established, the concentration data of different gas types in the gas concentration data are given a warning response, which also includes: Extract the warning response result of gas concentration, and extract the number of gas types corresponding to each warning level in the gas type data; Compare the number of gas types corresponding to each warning level with the corresponding preset number threshold; When the number of gas types in the first-level warning is lower than the corresponding number threshold, it is determined whether the number of gas types in the second-level warning and the third-level warning exceeds the corresponding number threshold; When the number of gas types in the third-level warning exceeds the corresponding number threshold, the detected gas data is marked as dangerous for gas concentration; When the number of gas types in the third-level warning does not exceed the corresponding number threshold, but the number of gas types in the second-level warning exceeds the corresponding number threshold, the gas concentration hazard judgment is performed by combining the gas concentration value corresponding to the gas type in the second-level warning with the gas concentration value corresponding to the gas type in the first-level warning; When the number of gas types in the third-level warning does not exceed the corresponding number threshold, and the number of gas types in the second-level warning does not exceed the corresponding number threshold, the gas concentration hazard judgment is made by using the gas concentration values ​​corresponding to the gas types in the third-level warning and the second-level warning combined with the gas concentration values ​​corresponding to the gas types in the first-level warning.

[0044] The technical effects of the above technical solution are: the grading standards of level 1 (normal concentration), level 2 (critical to abnormal), and level 3 (exceeding critical) reduce the false alarm rate by about 25%-35%. For example, the traditional single-threshold system may misjudge short-term fluctuations as danger, while the grading mechanism reduces such misjudgments through a two-level buffer. When the number of level 1 warning gas types is insufficient, the in-depth analysis of level 2 and level 3 gases is automatically triggered to avoid missing potential risks, and the false alarm rate is reduced by about 18%-22%. When the number of level 3 warning gases exceeds the standard, the danger is directly marked (no subsequent judgment is required), and the response time is shortened by about 40%-55%. Through the predefined judgment process, the amount of calculation is reduced by about 30%-40%. Only the gas data of the exceeding level is stored, and the storage space requirement is reduced by about 15%-25%. The number of gas types, concentration values ​​and warning levels are integrated to cover more potential risk scenarios. For example, although a gas has not reached the level 3 threshold, it fluctuates near the level 2 threshold for a long time, and the system can still determine the risk in combination with other gas data. The three-level early warning mechanism improves the recognition rate of sudden serious leaks by about 35%-45%, avoiding the delayed response caused by the high single threshold of the traditional system. The preset threshold can be adjusted according to the scenario, making the solution more stable in different environments and reducing the misjudgment rate by about 20%-30%. The modular design supports adding new gas types or adjusting thresholds without reconstructing the system, reducing the adaptation cost by about 40%-50%.

[0045] Specifically, the gas concentration value corresponding to the gas type of the second-level warning is combined with the gas concentration value corresponding to the gas type of the first-level warning to determine the gas concentration hazard, including: When the number of gas types in the third-level warning does not exceed the corresponding number threshold, but the number of gas types in the second-level warning exceeds the corresponding number threshold, the gas concentration value corresponding to the gas type in the first-level warning is retrieved; According to the gas concentration value corresponding to the gas type of the first-level warning, the concentration difference between the gas concentration value corresponding to the gas type of the first-level warning and the critical threshold is obtained; Retrieving the critical threshold, and performing ratio processing on the concentration difference between the gas concentration value corresponding to the gas type of the first-level warning and the critical threshold and the critical threshold, respectively, to obtain the gas concentration ratio corresponding to the gas type of the first-level warning as the first gas concentration ratio data; Retrieve the gas concentration value corresponding to the gas type of the second-level warning; According to the gas concentration value corresponding to the gas type of the second-level warning, the concentration difference between the gas concentration value corresponding to the gas type of the second-level warning and the abnormal threshold is obtained; Retrieving the abnormal threshold, and performing ratio processing on the concentration difference between the gas concentration value corresponding to the gas type of the secondary warning and the abnormal threshold and the abnormal threshold, respectively, to obtain the gas concentration ratio corresponding to the gas type of the secondary warning as the second gas concentration ratio data; Obtaining a first gas concentration coefficient by combining the first gas concentration ratio data with a gas concentration standard deviation of the second gas concentration ratio data; Wherein, the first gas concentration coefficient is obtained by the following formula: ;

[0046] Among them, R 01 represents the first gas concentration coefficient; Y 01 represents the average value of the gas concentration ratio corresponding to the first gas concentration ratio data; Y 01max represents the maximum value of the gas concentration ratio corresponding to the first gas concentration ratio data; δ represents the standard deviation of the gas concentration corresponding to the first gas concentration ratio data; σ represents the standard deviation of the gas concentration corresponding to the second gas concentration ratio data; Obtaining a second gas concentration coefficient using the second gas concentration ratio data; Wherein, the second gas concentration coefficient is obtained by the following formula: ;

[0047] Among them, R 02 represents the second gas concentration coefficient; Y 02 represents the average value of the gas concentration ratio corresponding to the second gas concentration ratio data; Y 02max represents the maximum value of the gas concentration ratio corresponding to the second gas concentration ratio data; δ represents the standard deviation of the gas concentration corresponding to the first gas concentration ratio data; σ represents the standard deviation of the gas concentration corresponding to the second gas concentration ratio data; comparing the first gas concentration coefficient and the second gas concentration coefficient; When the second gas concentration coefficient is not lower than the first gas concentration coefficient, the detected gas data is determined to be dangerous, and a gas concentration danger mark is performed on the detected gas data.

[0048] The technical effect of the above technical solution is: the first gas concentration coefficient obtained by the above technical solution is intended to comprehensively consider the average value, maximum value and related factors of the standard deviation of the first-level warning gas concentration ratio data, so as to measure the influence of the first-level warning gas concentration on the overall hazard judgment. The difference between the maximum value and the average value is divided by the standard deviation of the first gas concentration ratio data, which reflects the internal discreteness and relative changes of the first-level warning gas concentration ratio data, and then adjusted in combination with the relationship with the standard deviation of the second gas concentration ratio data. It represents the difference between the maximum value and the average value in the first gas concentration ratio data. The larger the difference, the greater the fluctuation of the first-level warning gas concentration between different types. It is then divided by the gas concentration standard deviation corresponding to the first gas concentration ratio data for normalization to reflect the degree of this fluctuation relative to its own standard deviation. The relative relationship between the standard deviation of the first gas concentration ratio data and the standard deviation of the second gas concentration ratio data is considered. If δ is larger than σ, it means that the first-level warning gas concentration data fluctuates more violently, and R 01 will increase accordingly. Overall, R 01 The larger the value is, the greater the impact of the comprehensive situation of the first-level warning gas concentration on the hazard judgment is. However, the specific situation still needs to be combined with R 02 Compare. with R 01 Similarly, the role of the secondary warning gas concentration situation on the overall hazard judgment is measured by combining the average value, maximum value, and the relationship with the standard deviation of the primary warning gas concentration ratio data. The difference between the maximum value and the average value is divided by the standard deviation of the second gas concentration ratio data to reflect the discrete characteristics within the secondary warning gas concentration ratio data, and then adjusted in combination with the relationship with the standard deviation of the first gas concentration ratio data. It reflects the difference between the maximum and average values ​​in the second gas concentration ratio data, indicating the degree of fluctuation of the concentration of the second-level warning gas between different types. After normalization by dividing by the standard deviation corresponding to the gas, it reflects the size of the fluctuation relative to its own standard deviation. Considering the relative relationship between the standard deviation of the second gas concentration ratio data and the standard deviation of the first gas concentration ratio data, if σ is larger than δ, it means that the fluctuation of the secondary warning gas concentration data is more obvious, R 02 Will increase. 02 The larger it is, the greater the impact of the comprehensive situation of the second-level warning gas concentration on the danger judgment.

[0049] When the second gas concentration coefficient R 02 Not less than the first gas concentration coefficient R 01 When the gas concentration of the second-level warning is greater than that of the first-level warning, it means that the comprehensive impact of the gas concentration of the second-level warning on the overall hazard determination is relatively greater or at least equivalent to that of the first-level warning. The number of secondary warning gas types has exceeded the corresponding threshold, indicating that the concentration of more gas types has reached a more serious warning level. 02 ≥R 01 It is further shown that the fluctuation of the gas concentration ratio data of the second-level warning, the relative degree of change, and the comprehensive relationship with the standard deviation of the data related to the first-level warning make its contribution to the hazard judgment greater or not less than that of the first-level warning. This shows that the gas concentration of the second-level warning is not only larger in quantity, but also more prominent in the characteristics of concentration change, and has reached or exceeded the dangerous level. Therefore, the detected gas data is judged to be dangerous at this time, and the gas concentration danger mark is carried out to remind relevant personnel to pay attention to potential gas hazards.

[0050] Combining the first-level (normal concentration) and second-level (critical to abnormal) gas concentration data, the degree of deviation between the concentration and the threshold is quantified through ratio analysis. Experiments show that compared with single-level judgment, the misjudgment rate is reduced by about 35%-45%. By using the ratio of the concentration difference to the threshold (rather than the absolute concentration value), the influence of the measurement unit or baseline difference is eliminated, the judgment standard is more unified, and the accuracy is improved by about 20%-30%. Complex calculations are triggered only when the second-level warning exceeds the standard, avoiding real-time analysis of the full amount of data, and reducing the amount of calculation by about 40%-55%. The critical threshold and abnormal threshold are pre-loaded, and the ratio calculation is optimized using the table lookup method, which shortens the response time by about 30%-40%. Through the first gas concentration coefficient (R 01 ) integrates standard deviations (δ and σ), automatically filters out noise data, and reduces storage requirements by about 18%-25%. Ratio calculation and standard deviation analysis can be processed in parallel, and the efficiency is improved by about 35%-45% on multi-core processors. At the same time, the average value of the concentration ratio (Y 01 , Y 02 ), maximum value (Y 01 max, Y 02 max) and standard deviation (δ, σ), covering more potential risk scenarios. For example, even if the concentration ratio of a gas is low but fluctuates violently (high σ), the system can still judge it as dangerous. The second gas concentration coefficient (R 02 ) through Y 02 Max strengthens extreme value monitoring, improving the recognition rate of sudden leaks by about 25%-35%. The preset threshold can be adjusted dynamically according to the scenario, making the solution more stable in different environments and reducing the false alarm rate by about 20%-30%. The modular design supports adding new gas types or adjusting thresholds without reconstructing the system, reducing the adaptation cost by about 40%-50%.

[0051] Specifically, the gas concentration hazard determination is performed using the gas concentration values ​​corresponding to the gas types of the third-level warning and the second-level warning combined with the gas concentration values ​​corresponding to the gas types of the first-level warning, including: When the number of gas types in the third-level warning does not exceed the corresponding number threshold, and the number of gas types in the second-level warning does not exceed the corresponding number threshold, the difference between the gas concentration value corresponding to the gas type in the first-level warning and the abnormal threshold is processed with the critical threshold to obtain the ratio data between the difference between the gas concentration value corresponding to the gas type in the first-level warning and the critical threshold and the abnormal threshold as the third gas concentration ratio data; Obtaining a third gas concentration coefficient using the third gas concentration ratio data; The third gas concentration coefficient is obtained by the following formula: ;

[0052] Among them, R 03 Indicates the third gas concentration coefficient; Y 01 represents the average value of the gas concentration ratio corresponding to the first gas concentration ratio data; Y 01max represents the maximum value of the gas concentration ratio corresponding to the first gas concentration ratio data; Y 01min represents the minimum value of the gas concentration ratio corresponding to the first gas concentration ratio data; δ represents the standard deviation of the gas concentration corresponding to the first gas concentration ratio data; σ represents the standard deviation of the gas concentration corresponding to the second gas concentration ratio data; Performing ratio processing on the difference between the gas concentration value corresponding to the gas type of the secondary warning and the abnormal threshold and the abnormal threshold, obtaining the ratio data between the difference between the gas concentration value corresponding to the gas type of the secondary warning and the abnormal threshold and the abnormal threshold as the second gas concentration ratio data; Performing ratio processing on the difference between the gas concentration value corresponding to the gas type of the third-level warning and the abnormal threshold and the abnormal threshold, obtaining the ratio data between the difference between the gas concentration value corresponding to the gas type of the third-level warning and the abnormal threshold and the abnormal threshold as the third gas concentration ratio data; Obtaining a comprehensive gas concentration coefficient using the second gas concentration ratio data and the third gas concentration ratio data; The comprehensive gas concentration coefficient is obtained by the following formula: ;

[0053] Among them, R z Indicates the comprehensive gas concentration coefficient; Y 02 represents the average value of the gas concentration ratio corresponding to the second gas concentration ratio data; Y 02max represents the maximum value of the gas concentration ratio corresponding to the second gas concentration ratio data; δ represents the standard deviation of the gas concentration corresponding to the first gas concentration ratio data; σ represents the standard deviation of the gas concentration corresponding to the second gas concentration ratio data; Y 03represents the average value of the gas concentration ratio corresponding to the third gas concentration ratio data; Y 03max represents the maximum value of the gas concentration ratio corresponding to the third gas concentration ratio data; μ represents the standard deviation of the gas concentration corresponding to the third gas concentration ratio data; comparing the third gas concentration coefficient with the comprehensive gas concentration coefficient; When the comprehensive gas concentration coefficient is not lower than the third gas concentration coefficient, the detected gas data is determined to be dangerous, and a gas concentration danger mark is performed on the detected gas data.

[0054] The technical effect of the above technical solution is: the third gas concentration coefficient obtained in the above technical solution is mainly constructed based on the characteristics of the gas concentration ratio data of the first-level warning. By considering the average value, maximum value, minimum value of the first-level warning gas concentration ratio and the relationship with the standard deviation of the first and second-level warning gas concentration ratio data, the impact of the first-level warning gas concentration on the overall hazard judgment is comprehensively measured. The difference between the maximum value and the average value and the difference between the average value and the minimum value are used, combined with the standard deviation, to make adjustments to reflect the fluctuation and discreteness of the data.

[0055] Indicates the difference between the average value of the first gas concentration ratio data and the average value of the maximum and minimum values. The larger the difference, the more uneven the distribution of the first-level warning gas concentration among different types, and the greater the fluctuation. Divide by δ for normalization to reflect the degree of this difference relative to its own standard deviation. Considering the relative relationship between the standard deviation of the first gas concentration ratio data and the standard deviation of the second gas concentration ratio data, if δ is larger than σ, it means that the concentration data of the first-level warning gas fluctuates more violently, which will make R 03 Therefore, R 03 The larger the value is, the greater the impact of the fluctuation of the first-level warning gas concentration on the danger assessment.

[0056] In the process of obtaining the comprehensive gas concentration coefficient Rz, the above technical solution combines the gas concentration ratio data characteristics of the second-level warning and the third-level warning, and considers the relationship with the standard deviation of the first-level warning gas concentration ratio data to construct a coefficient that comprehensively reflects the impact of the second- and third-level warning gas concentrations on the hazard judgment. The difference between the maximum and average values ​​of the second-level and third-level warning gas concentration ratio data is calculated respectively, and adjusted in combination with their respective standard deviations and the relationship with the standard deviations related to other levels of warning, and finally the two are multiplied.

[0057] For the second level warning part, It represents the difference between the maximum and average values ​​of the second gas concentration ratio data, reflecting the fluctuation degree of the concentration of the second-level warning gas between different types. After normalization by dividing by σ, it is adjusted in combination with the relationship with the standard deviation δ of the first gas concentration ratio data. Similarly, for the third-level warning part, It reflects the fluctuation of the third-level warning gas concentration and its relationship with the standard deviation of itself and the first-level warning. The larger the Rz obtained by multiplying the two, the greater the impact of the comprehensive fluctuation of the second- and third-level warning gas concentrations on the danger judgment.

[0058] When the comprehensive gas concentration coefficient Rz is not lower than the third gas concentration coefficient R 03 This means that the comprehensive impact of the second and third level warning gas concentrations on the overall hazard assessment is relatively greater or at least equivalent to the first level warning. From a physical perspective: Although the number of gas types in the third and second level warnings does not exceed the corresponding threshold, Rz≥R 03 It shows that the fluctuation of the gas concentration ratio data of the second and third level warnings, the relative degree of change, and the comprehensive relationship with the standard deviation of the data related to the first level warning make their contribution to the hazard judgment greater or not less than that of the first level warning. This means that the gas concentration of the second and third level warnings is more prominent in terms of concentration change characteristics. Even if the quantity does not exceed the threshold, its concentration fluctuation and change have reached or exceeded the dangerous level. Because the second and third level warnings themselves represent more serious warning levels, when their combined impact on the hazard judgment is not less than that of the first level warning, it means that the potential gas hazard risk is high, so at this time, the detected gas data is judged to be dangerous, and the gas concentration hazard mark is made to remind relevant personnel to take precautions.

[0059] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0060] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

Claims

1. A method for identifying gas components and monitoring concentration in transformer oil based on spectral analysis, characterized in that: The steps include: S1: Sample processing and collection: inject the oil sample in the transformer into the degassing device for degassing, collect the gas after degassing, and perform equipment inspection on the degassing device before degassing, and mark the collected gas as target detection gas data; S2: Spectral data acquisition and processing: The target detected gas data is collected by using an infrared spectrometer for gas molecule data acquisition, and the collected gas molecule data is preprocessed to obtain gas spectrum data; S3: Spectral data feature analysis: extracting feature data from gas spectral data, performing feature screening on the extracted feature data, generating a gas feature vector after feature screening, and marking the generated gas feature vector as gas feature analysis data; When extracting characteristic data from the gas spectrum data in S3, the extracted characteristic data include peak area, peak height, peak width and peak shape.

2. The method for identifying gas components and monitoring concentration in transformer oil based on spectral analysis according to claim 1, characterized in that: In S1, the oil sample in the transformer is injected into the degassing device for degassing, the gas after degassing is collected, and the degassing device is tested before degassing, including: The degassing device is a vacuum degassing device. Before degassing the oil sample, the degassing device shall be inspected. The equipment inspection includes appearance inspection, electrical inspection, vacuum system inspection, heating inspection, filtration inspection and safety inspection. After the equipment is tested and qualified, use the sampling equipment to take the oil sample from the transformer and inject the taken oil sample into the sample chamber of the degassing device; After the oil sample is injected, the working parameters of the degassing device are set. The working parameters include working time and temperature. The temperature is between 40°C and 60°C, and the time is 30 minutes to 1 hour. The gas released during the degassing process is transferred to the gas collection bottle through a conduit; The gas in the gas collection bottle is marked as the target detection gas data.

3. The method for identifying gas components and monitoring concentration in transformer oil based on spectral analysis according to claim 2, characterized in that: In S2, the target gas detection data is collected by using an infrared spectrometer to collect gas molecule data, and the collected gas molecule data is preprocessed, including: Before the infrared spectrometer collects gas molecule data, it is first calibrated, including optical system calibration, background calibration and wavelength calibration; injecting the collected target detection gas data into the sample chamber of the infrared spectrometer; After the injection is completed, the infrared spectrometer is started to scan the target detection gas data in the sample chamber, and the scanning includes background scanning, sample scanning and multiple scanning; Background scanning is to record the spectral data when there is no target detection gas data; sample scanning is to scan the target detection gas data; multiple scanning is to scan the target detection gas data multiple times and take the average value; After the target detection gas data in the sample chamber is scanned, the gas molecule data of the target detection gas data is obtained; The gas molecule data are preprocessed, including baseline correction, spectrum normalization, denoising and data smoothing; After data preprocessing, gas spectrum data is obtained.

4. The method for identifying gas components and monitoring concentration in transformer oil based on spectral analysis according to claim 3, characterized in that: In S3, feature data is extracted from gas spectrum data, and feature screening is performed on the extracted feature data. After feature screening, a gas feature vector is generated. The generated gas feature vector includes: Peak shape is to analyze the shape of the peak in the gas spectrum data; peak height is to measure the height of each absorption peak in the gas spectrum data; peak width is to analyze the width of the peak in the gas spectrum data; The peak area is calculated based on the peak shape, including the triangle method and the numerical integration method. The triangle method is used for the calculation of symmetrical peaks, and the numerical integration method is used for the calculation of asymmetrical peaks. The calculation formula of the triangle method is as follows: ； The calculation formula of numerical integration method is as follows: ； in, It is expressed as the width of each cell; Expressed as The function value of sampling points; It is expressed as the total number of sampling points; Perform feature screening on the extracted feature data, which includes correlation analysis and principal component analysis; The data after feature screening are combined into a feature vector, each feature vector represents a spectrum sample, and each spectrum sample is standardized to obtain a gas feature vector after the standardization process.

5. The method for identifying gas components and monitoring concentration in transformer oil based on spectral analysis according to claim 4, characterized in that: include: S4: Gas identification and collation: Identify the gas type of the gas characteristic analysis data according to the gas identification model, and generate visualization data for the identified gas type. After the visualization data is generated, the gas type data is obtained; Divide the gas feature vectors into training set and test set; Retrieving a gas recognition model from a database, wherein the gas recognition model is a deep learning model; The gas recognition model is trained using the training set of gas feature vectors; The model training process is to input the feature vectors in the training set into the gas recognition model, calculate the loss function after the feature vector input is completed, and use the k-fold cross-validation method to verify the gas recognition model after the loss function calculation, and complete the model training after verification; Use the test set to evaluate the performance of the trained gas recognition model. The performance evaluation includes accuracy evaluation, recall evaluation, F1 score evaluation, and confusion matrix evaluation. The gas feature vector is input into the gas identification model after performance evaluation to predict the type of gas, and the confidence of each gas type is obtained after the type prediction; Visualize the confidence of each gas type, including bar charts, scatter plots and boundary plots; Generate a result report using the visualized converted data, and annotate the generated result report as gas type data.

6. The method for identifying gas components and monitoring concentration in transformer oil based on spectral analysis according to claim 5, characterized in that: include: S5: quantitative concentration analysis: retrieve gas concentration data from gas type data, and perform sample concentration analysis on the retrieved concentration data to obtain gas concentration data; the gas type data includes the type name, concentration value, unit and detection time of each gas type, and the concentration value is obtained by concentration value detection through sample scanning of the infrared spectrometer; Perform quantitative analysis based on the obtained concentration values; The quantitative analysis process includes: The standard concentration data of each gas type is retrieved from the database, and a standard curve is established for the standard concentration data; Establish a detection curve using the acquired concentration values; Overlapping and comparing the established standard curve with the established detection curve; According to the curve overlap comparison results, determine whether the test curve overlaps with the standard curve; If the detection curve is lower than the standard curve, the concentration value corresponding to the detection curve is within the normal range of the gas type; If the detection curve coincides with the standard curve, the concentration value corresponding to the detection curve is the critical range of the gas type; If the detection curve exceeds the standard curve, the concentration value corresponding to the detection curve is the abnormal range of the gas type; Finally, the curve overlap comparison results are marked as gas concentration data.

7. The method for identifying gas components and monitoring concentration in transformer oil based on spectral analysis according to claim 6, characterized in that: include: S6: Gas concentration judgment and warning: formulate different concentration warning ranges for different gas types in the gas concentration data, and make warning responses to the concentration data of different gas types in the gas concentration data according to the formulated different concentration warning ranges; According to the curve overlap comparison results of the gas concentration data, the analysis results of each gas type and the concentration value corresponding to the gas type are obtained; The analysis results include normal concentration, critical concentration and abnormal concentration; Formulate an early warning range for each gas type, and the early warning range is formulated by setting critical thresholds and abnormal thresholds for each gas type; The critical concentration and abnormal concentration data in the analysis results are used for early warning response according to the established early warning range; The early warning response process includes: Confirm the gas type data in the critical concentration and abnormal concentration, and integrate the confirmed gas type data and the concentration value of the data; The integrated data is divided into warning levels according to the critical threshold and abnormal threshold in the established warning range; The warning levels include level one, level two and level three; Different intensities of early warning will be issued according to different warning levels.

8. The method for identifying gas components and monitoring concentration in transformer oil based on spectral analysis according to claim 7, characterized in that: According to the different concentration warning ranges established, the concentration data of different gas types in the gas concentration data are warned and responded to, including: Extract the warning response result of gas concentration, and extract the number of gas types corresponding to each warning level in the gas type data; Compare the number of gas types corresponding to each warning level with the corresponding preset number threshold; When the number of gas types in the first-level warning is lower than the corresponding number threshold, it is determined whether the number of gas types in the second-level warning and the third-level warning exceeds the corresponding number threshold; When the number of gas types in the third-level warning exceeds the corresponding number threshold, the detected gas data is marked as dangerous for gas concentration; When the number of gas types in the third-level warning does not exceed the corresponding number threshold, but the number of gas types in the second-level warning exceeds the corresponding number threshold, the gas concentration hazard judgment is performed by combining the gas concentration value corresponding to the gas type in the second-level warning with the gas concentration value corresponding to the gas type in the first-level warning; When the number of gas types in the third-level warning does not exceed the corresponding number threshold, and the number of gas types in the second-level warning does not exceed the corresponding number threshold, the gas concentration hazard judgment is made by using the gas concentration values ​​corresponding to the gas types in the third-level warning and the second-level warning combined with the gas concentration values ​​corresponding to the gas types in the first-level warning.

9. The method for identifying gas components and monitoring concentration in transformer oil based on spectral analysis according to claim 8, characterized in that: The gas concentration value corresponding to the gas type of the second-level warning is combined with the gas concentration value corresponding to the gas type of the first-level warning to determine the gas concentration hazard, including: When the number of gas types in the third-level warning does not exceed the corresponding number threshold, but the number of gas types in the second-level warning exceeds the corresponding number threshold, the gas concentration value corresponding to the gas type in the first-level warning is retrieved; According to the gas concentration value corresponding to the gas type of the first-level warning, the concentration difference between the gas concentration value corresponding to the gas type of the first-level warning and the critical threshold is obtained; Retrieving the critical threshold, and performing ratio processing on the concentration difference between the gas concentration value corresponding to the gas type of the first-level warning and the critical threshold and the critical threshold, respectively, to obtain the gas concentration ratio corresponding to the gas type of the first-level warning as the first gas concentration ratio data; Retrieve the gas concentration value corresponding to the gas type of the second-level warning; According to the gas concentration value corresponding to the gas type of the second-level warning, the concentration difference between the gas concentration value corresponding to the gas type of the second-level warning and the abnormal threshold is obtained; Retrieving the abnormal threshold, and performing ratio processing on the concentration difference between the gas concentration value corresponding to the gas type of the secondary warning and the abnormal threshold and the abnormal threshold, respectively, to obtain the gas concentration ratio corresponding to the gas type of the secondary warning as the second gas concentration ratio data; Obtaining a first gas concentration coefficient by combining the first gas concentration ratio data with a gas concentration standard deviation of the second gas concentration ratio data; Obtaining a second gas concentration coefficient using the second gas concentration ratio data; comparing the first gas concentration coefficient and the second gas concentration coefficient; When the second gas concentration coefficient is not lower than the first gas concentration coefficient, the detected gas data is determined to be dangerous, and a gas concentration danger mark is performed on the detected gas data.

10. The method for identifying gas components and monitoring concentration in transformer oil based on spectral analysis according to claim 9, characterized in that: The gas concentration hazard judgment is made by using the gas concentration values ​​corresponding to the gas types of the third-level warning and the second-level warning and the gas concentration values ​​corresponding to the gas types of the first-level warning, including: When the number of gas types in the third-level warning does not exceed the corresponding number threshold, and the number of gas types in the second-level warning does not exceed the corresponding number threshold, the difference between the gas concentration value corresponding to the gas type in the first-level warning and the abnormal threshold is processed with the critical threshold to obtain the ratio data between the difference between the gas concentration value corresponding to the gas type in the first-level warning and the critical threshold and the abnormal threshold as the third gas concentration ratio data; Obtaining a third gas concentration coefficient using the third gas concentration ratio data; Performing ratio processing on the difference between the gas concentration value corresponding to the gas type of the secondary warning and the abnormal threshold and the abnormal threshold, obtaining the ratio data between the difference between the gas concentration value corresponding to the gas type of the secondary warning and the abnormal threshold and the abnormal threshold as the second gas concentration ratio data; Performing ratio processing on the difference between the gas concentration value corresponding to the gas type of the third-level warning and the abnormal threshold and the abnormal threshold, obtaining the ratio data between the difference between the gas concentration value corresponding to the gas type of the third-level warning and the abnormal threshold and the abnormal threshold as the third gas concentration ratio data; Obtaining a comprehensive gas concentration coefficient using the second gas concentration ratio data and the third gas concentration ratio data; comparing the third gas concentration coefficient with the comprehensive gas concentration coefficient; When the comprehensive gas concentration coefficient is not lower than the third gas concentration coefficient, the detected gas data is determined to be dangerous, and a gas concentration danger mark is performed on the detected gas data.

Images