A method for detecting gas in transformer oil based on laser spectroscopy

Abstract

This invention discloses a method for detecting gas in transformer oil based on laser spectroscopy, belonging to the field of gas detection technology. The method includes the following steps: S1: Collecting real-time data on gas in transformer oil, processing the real-time data, and determining the characteristic data of the gas in the transformer oil; S2: Evaluating and optimizing the fault identification model for gas detection in transformer oil to determine the optimal model; S3: Performing predictive analysis on the characteristic data of the gas in transformer oil. This invention solves the problems of existing methods that cannot accurately detect gas in transformer oil, cannot effectively predict the nature and type of transformer faults, and cannot ensure long-term stable operation of transformers, resulting in poor transformer performance. This invention can accurately detect gas in transformer oil, effectively predict the nature and type of transformer faults, ensure long-term stable operation of transformers, and improve transformer performance.

Description

Technical Field

[0001] This invention relates to the field of gas detection technology, specifically a method for detecting gas in transformer oil based on laser spectroscopy. Background Technology

[0002] Transformers are the backbone of a power system and the foundation for ensuring reliable power supply. A problem with even one major transformer can cause local and system-wide system shutdowns, leading to power outages. Therefore, the safe and reliable operation of transformers is of paramount importance. During operation, transformers may experience malfunctions that cause abnormal changes in the gas bubbles in the oil. These bubbles contain various gases, such as… The generation of these gases is often related to the deterioration of insulation materials, and if they are not detected and treated in time, they may cause accidents.

[0003] Chinese patent CN118465159A discloses a transformer oil testing box, comprising: a box body with a receiving cavity; an oil inlet pipe passing through the side wall of the box body; a heat insulation structure disposed within the receiving cavity, the heat insulation structure including a heating seat, a heat-conducting cylinder disposed above the heating seat, and a heat-conducting pipe disposed within the heat-conducting cylinder, the heating seat having a heating cavity, the heat-conducting pipe communicating with the heating cavity, the heating cavity containing a heat-conducting fluid that can flow from the heating cavity into the heat-conducting pipe; an oil chromatography column comprising a first connecting section, a heat-conducting section, and a second connecting section arranged sequentially, the end of the first connecting section away from the heat-conducting section communicating with the oil inlet pipe, the heat-conducting section being disposed on the outside of the heat-conducting cylinder; and a detector disposed within the receiving cavity, the end of the second connecting section away from the heat-conducting section communicating with the detector; this patent can effectively solve the problem of high energy consumption in constant temperature chambers in related technologies; however, this patent has the following drawbacks:

[0004] The existing methods cannot accurately detect gas in transformer oil, cannot effectively predict the nature and type of transformer faults, and cannot ensure long-term stable operation of transformers, resulting in poor transformer performance. Summary of the Invention

[0005] The purpose of this invention is to provide a method for detecting gas in transformer oil based on laser spectroscopy, which can accurately detect gas in transformer oil, effectively predict the nature and type of transformer faults, ensure long-term stable operation of transformers, improve transformer performance, and solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for detecting gas in transformer oil based on laser spectroscopy includes the following steps:

[0008] S1: Collect real-time data of gas in transformer oil based on laser spectroscopy, process the real-time data of gas in transformer oil based on laser spectroscopy, and determine the characteristic data of gas in transformer oil based on laser spectroscopy.

[0009] S2: Based on the requirements for gas detection in transformer oil based on laser spectroscopy, a fault identification model for gas detection in transformer oil based on laser spectroscopy is constructed. The fault identification model for gas detection in transformer oil based on laser spectroscopy is evaluated and optimized to determine the optimal fault identification model for gas detection in transformer oil based on laser spectroscopy.

[0010] S3: Based on the optimal fault identification model for gas detection in transformer oil based on laser spectroscopy, predictive analysis is performed on the characteristic data of gas in transformer oil based on laser spectroscopy to determine the fault identification result for gas detection in transformer oil based on laser spectroscopy.

[0011] S4: Develop an intelligent management and control scheme for gas detection in transformer oil based on laser spectroscopy. Based on this scheme, implement intelligent management and control of gas detection in transformer oil based on laser spectroscopy to ensure long-term stable operation of the transformer.

[0012] Preferably, in step S1, real-time data on gas in transformer oil based on laser spectroscopy is collected, and the following operations are performed:

[0013] The transformer oil sample is determined by taking a sample directly from the oil sampling port on the transformer.

[0014] A constant-temperature vacuum degassing technique was used to efficiently degas transformer oil samples.

[0015] The gas dissolved in the transformer oil sample is released, while preventing the transformer oil sample from contacting the air. The degassed gas is then introduced into the photoacoustic cell.

[0016] A laser of a specific wavelength is emitted to irradiate the gas in the photoacoustic cell. After the gas molecules absorb the laser energy, their temperature rises and they release heat energy, which in turn generates pressure waves to determine the gas concentration.

[0017] Obtain gas concentration, trends, gas ratios, and gas change rates;

[0018] Real-time data on gas in transformer oil based on laser spectroscopy were determined.

[0019] Preferably, in step S1, the real-time data on gas in transformer oil based on laser spectroscopy is processed by performing the following operations:

[0020] Acquire real-time data on gas in transformer oil based on laser spectroscopy;

[0021] Cleaning and processing of real-time data on gases in transformer oil based on laser spectroscopy, including:

[0022] Consistency checks were performed on real-time data of gas in transformer oil based on laser spectroscopy.

[0023] Based on the reasonable value range and interrelationship of each variable in the real-time data of gas in transformer oil based on laser spectroscopy, check whether the real-time data of gas in transformer oil based on laser spectroscopy meets the requirements.

[0024] Remove inconsistent data from real-time data on gas in transformer oil based on laser spectroscopy that are outside the normal range, logically unreasonable, or contradictory.

[0025] Invalid and missing values ​​were processed in real-time data of gas in transformer oil based on laser spectroscopy.

[0026] Remove invalid and missing data that are not valuable for the detection of gas in transformer oil based on laser spectroscopy from the real-time data of gas in transformer oil based on laser spectroscopy;

[0027] Real-time data on gas in transformer oil based on laser spectroscopy were identified as valuable for gas detection in transformer oil.

[0028] Preferably, real-time data on gases in transformer oil based on laser spectroscopy, which are valuable for detecting gases in transformer oil, are identified, including:

[0029] Real-time monitoring removes invalid data that is of no value to the detection of gases in transformer oil based on laser spectroscopy;

[0030] Extract invalid data that has already been removed;

[0031] Sort the invalid data according to the generation time of each invalid data, and obtain the data set corresponding to the invalid data;

[0032] Retrieve a preset time period; wherein the preset time period ranges from 3 min to 7 min.

[0033] The data set corresponding to the invalid data is divided according to the preset unit time period to obtain multiple invalid data groups;

[0034] The invalid evaluation coefficient for each invalid data group is obtained by utilizing the data generation time interval of the invalid data contained in the invalid data group; wherein, the invalid evaluation coefficient is obtained by the following formula:

[0035] ;

[0036] Where P represents the invalidity evaluation coefficient; n represents the number of invalid data points contained in each invalid data group; T c T represents the duration of a unit of time; i T represents the time of generation corresponding to the i-th invalid data. i-1 Indicates the generation time corresponding to the (i-1)th invalid data; w i w represents the weight value corresponding to the i-th invalid data; i-1 G represents the weight value corresponding to the (i-1)th invalid data point; i G represents the time interval between the generation time of the most recent valid data with the same weight value corresponding to the i-th invalid data; i-1 G represents the time interval between the generation time of the most recent valid data with the same weight value corresponding to the (i-1)th invalid data and the i-th invalid data; p This represents the average of the time intervals between the generation of n invalid data points.

[0037] The invalid evaluation coefficient is compared with a preset invalid evaluation coefficient threshold;

[0038] Extract invalid evaluation coefficients that exceed the invalid evaluation coefficient threshold and use them as target invalid evaluation coefficients;

[0039] The invalid data is filled using the target invalidity evaluation coefficient.

[0040] Preferably, filling invalid data with the target invalidity evaluation coefficient includes:

[0041] Extract the invalid data group to which each target invalid evaluation coefficient belongs, and use it as the target invalid data group;

[0042] Extract the generation time corresponding to each invalid data point in the invalid data group to which each invalid evaluation coefficient of the target belongs;

[0043] Based on the generation time of each invalid data, obtain the valid data that comes before the generation time of each invalid data and the valid data that comes after the generation time of each invalid data;

[0044] By combining the valid data that occurred earlier and the valid data that occurred later in time with the time adjacent to the time each invalid data was generated, and the target invalid evaluation coefficient, the invalid data compensation coefficient corresponding to the target invalid data group is obtained.

[0045] The invalid data compensation coefficient corresponding to the target invalid data group is obtained by the following formula:

[0046] ;

[0047] Where B represents the invalid data compensation coefficient corresponding to the target invalid data group; X 01 and X 02 These represent the data values ​​of the valid data occurring before and after the time of each invalid data occurrence; X p P represents the average value of valid data; P represents the invalid evaluation coefficient; P0 represents the preset invalid evaluation coefficient threshold; X z This represents the intermediate value of the valid data.

[0048] The replacement data corresponding to each invalid data in the target invalid data group is obtained using the invalid data compensation coefficient corresponding to the target invalid data group;

[0049] Replace each invalid data in the target invalid data group with the replacement data corresponding to each invalid data in the target invalid data group.

[0050] Preferably, obtaining replacement data corresponding to each invalid data in the target invalid data group using the invalid data compensation coefficient corresponding to the target invalid data group includes:

[0051] Retrieve the invalid data compensation coefficients corresponding to the target invalid data group;

[0052] Retrieve valid data from each invalid data group in the target invalid data group, including valid data with the earliest and latest times that are adjacent to each invalid data point.

[0053] By utilizing the valid data that occurs earlier and the valid data that occurs later in the time adjacent to the time of each invalid data in the target invalid data group, the data baseline value corresponding to each invalid data in the target invalid data group is obtained;

[0054] The data baseline value is obtained using the following formula:

[0055] ;

[0056] Among them, X k This represents the baseline value corresponding to each invalid data point; X 01 and X 02 These represent the data values ​​of the valid data occurring earlier and later in the time frame adjacent to the time each invalid data point was generated; P0 represents the preset invalid evaluation coefficient threshold; P z This represents the median value of the invalid evaluation coefficients corresponding to all invalid data groups;

[0057] The replacement data for each invalid data is obtained by combining the invalid data compensation coefficient corresponding to the target invalid data group with the data baseline value corresponding to each invalid data in the target invalid data group;

[0058] The replacement data corresponding to each invalid data is obtained by the following formula:

[0059] ;

[0060] Among them, X t This represents the replacement data for each invalid data; X k This represents the baseline value corresponding to each invalid data point; X 01 and X 02 represents the data values ​​of the valid data that occurred earlier and later in the time frame adjacent to the time each invalid data was generated; B represents the invalid data compensation coefficient corresponding to the target invalid data group.

[0061] Preferably, in step S1, the real-time data on gas in transformer oil based on laser spectroscopy is processed, and the following operations are also performed:

[0062] To acquire real-time data on gas in transformer oil based on laser spectroscopy, which is valuable for gas detection in transformer oil.

[0063] The real-time data of gas in transformer oil based on laser spectroscopy, which is valuable for gas detection in transformer oil after cleaning, are normalized.

[0064] Eliminate dimensional differences between real-time data on gases in transformer oil based on laser spectroscopy;

[0065] Real-time data on gas in transformer oil based on laser spectroscopy were determined using standardized data.

[0066] Feature extraction was performed on real-time data of gas in transformer oil based on laser spectroscopy, which was standardized by data.

[0067] Extract features that are valuable for detecting gases in transformer oil based on laser spectroscopy;

[0068] Characteristic data of gas in transformer oil based on laser spectroscopy were determined.

[0069] Preferably, in step S2, a fault identification model for detecting gas in transformer oil based on laser spectroscopy is constructed, and the fault identification model for detecting gas in transformer oil based on laser spectroscopy is evaluated and optimized by performing the following operations:

[0070] Based on the requirements for gas detection in transformer oil using laser spectroscopy, historical gas concentrations, trends, gas ratios, and gas change rates are collected to determine historical gas data in transformer oil based on laser spectroscopy.

[0071] Historical data on gases in transformer oil based on laser spectroscopy are segmented;

[0072] Determine the gas detection training set and the gas detection test set;

[0073] Based on the requirements for gas detection in transformer oil using laser spectroscopy, a neural network model framework suitable for gas detection in transformer oil using laser spectroscopy is selected.

[0074] Based on the gas detection training set, a selected neural network model framework suitable for gas detection in transformer oil based on laser spectroscopy was trained.

[0075] A fault identification model for gas detection in transformer oil based on laser spectroscopy was determined;

[0076] Based on a gas detection test set, the performance of a fault identification model for gas detection in transformer oil based on laser spectroscopy is evaluated.

[0077] The performance test evaluation results of the fault identification model for gas detection in transformer oil based on laser spectroscopy were determined.

[0078] The performance test and evaluation results of the fault identification model for gas detection in transformer oil based on laser spectroscopy are analyzed.

[0079] A parameter adjustment and optimization scheme for a fault identification model for gas detection in transformer oil based on laser spectroscopy was determined.

[0080] The parameters of the fault identification model for gas detection in transformer oil based on laser spectroscopy are adjusted and optimized according to the parameter adjustment and optimization scheme.

[0081] The optimal fault identification model for gas detection in transformer oil based on laser spectroscopy was determined.

[0082] Preferably, in step S3, the predictive analysis of the gas characteristic data in transformer oil based on laser spectroscopy is performed by performing the following operations:

[0083] Acquire gas characteristic data in transformer oil based on laser spectroscopy;

[0084] The characteristic data of gas in transformer oil based on laser spectroscopy are input into the optimal fault identification model for gas detection in transformer oil based on laser spectroscopy.

[0085] Based on the optimal fault identification model for gas detection in transformer oil based on laser spectroscopy, predictive analysis is performed on the characteristic data of gas in transformer oil based on laser spectroscopy.

[0086] The results of fault identification based on laser spectroscopy for gas detection in transformer oil were determined.

[0087] Preferably, in step S4, intelligent control is performed on the detection of gas in transformer oil based on laser spectroscopy, and the following operations are performed:

[0088] Obtain fault identification results for gas detection in transformer oil based on laser spectroscopy;

[0089] Analysis of fault identification results based on gas detection in transformer oil using laser spectroscopy;

[0090] An intelligent control scheme for gas detection in transformer oil based on laser spectroscopy was determined;

[0091] Intelligent management and control of gas detection in transformer oil based on laser spectroscopy is implemented.

[0092] Specifically, based on the fault identification results of gas detection in transformer oil based on laser spectroscopy, the nature and type of transformer fault are determined, timely early warning is issued, and intelligent control measures are formulated. Transformer faults are repaired in a timely manner, and the transformer is regularly maintained and optimized according to the actual operating conditions to ensure long-term stable operation of the transformer.

[0093] Compared with the prior art, the beneficial effects of the present invention are:

[0094] 1. This invention directly samples transformer oil through the oil sampling port on the transformer to determine the transformer oil sample. It uses constant temperature vacuum degassing technology to efficiently degas the transformer oil sample, releasing the gas dissolved in the transformer oil sample while avoiding contact between the transformer oil sample and air. The degassed gas is introduced into a photoacoustic cell, and a laser of a specific wavelength is emitted to irradiate the gas in the photoacoustic cell. After the gas molecules absorb the laser energy, their temperature rises and they release heat energy, thereby generating pressure waves. The gas concentration is determined, and the gas concentration, trend, gas ratio, and gas change rate are obtained to determine real-time data of gas in transformer oil based on laser spectrum.

[0095] 2. This invention processes real-time data on gases in transformer oil based on laser spectroscopy to determine the characteristic data of these gases. Based on the detection requirements for gases in transformer oil based on laser spectroscopy, a fault identification model for gas detection in transformer oil based on laser spectroscopy is constructed. This model is then evaluated and optimized to determine the optimal model. Based on this optimal model, predictive analysis is performed on the characteristic data of gases in transformer oil based on laser spectroscopy to determine the fault identification results. An intelligent management and control scheme for gas detection in transformer oil based on laser spectroscopy is then formulated. This scheme enables intelligent management and control of gas detection in transformer oil based on laser spectroscopy, allowing for accurate detection of gases in the transformer oil, effective prediction of the nature and type of transformer faults, ensuring long-term stable operation of the transformer, and improving its performance. Attached Figure Description

[0096] Figure 1 This is a flowchart of the gas detection method in transformer oil based on laser spectroscopy according to the present invention. Detailed Implementation

[0097] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0098] To address the current limitations in accurately detecting gases in transformer oil, effectively predicting the nature and type of transformer faults, and ensuring long-term stable operation, which leads to poor transformer performance, please refer to [link to relevant documentation]. Figure 1 This embodiment provides the following technical solution:

[0099] A method for detecting gas in transformer oil based on laser spectroscopy includes the following steps:

[0100] S1: Collect real-time data of gas in transformer oil based on laser spectroscopy, process the real-time data of gas in transformer oil based on laser spectroscopy, and determine the characteristic data of gas in transformer oil based on laser spectroscopy.

[0101] In this embodiment, real-time data on gas in transformer oil based on laser spectroscopy is collected, and the following operations are performed:

[0102] The transformer oil sample is determined by taking a sample directly from the oil sampling port on the transformer.

[0103] A constant-temperature vacuum degassing technique was used to efficiently degas transformer oil samples.

[0104] The gas dissolved in the transformer oil sample is released, while preventing the transformer oil sample from contacting the air. The degassed gas is then introduced into the photoacoustic cell.

[0105] It should be noted that constant temperature vacuum degassing technology is an advanced process that improves gas purity by performing vacuum degassing under constant temperature and pressure. It has advantages such as energy saving, environmental protection, and no pollution, and is widely used in industrial production, scientific research experiments and other fields.

[0106] In isothermal vacuum degassing technology, gas is drawn into a vacuum device and degassed at a pressure lower than atmospheric pressure. Compared with traditional vacuum degassing technology, isothermal vacuum degassing technology has higher desorption efficiency and lower energy consumption. In practical applications, this technology can effectively reduce energy consumption in the gas treatment process, reduce environmental pollution, lower operating costs, and improve product performance.

[0107] A laser of a specific wavelength is emitted to irradiate the gas in the photoacoustic cell. After the gas molecules absorb the laser energy, their temperature rises and they release heat energy, which in turn generates pressure waves to determine the gas concentration.

[0108] Obtain gas concentration, trends, gas ratios, and gas change rates;

[0109] Real-time data on gas in transformer oil based on laser spectroscopy were determined.

[0110] In this embodiment, the real-time data of gas in transformer oil based on laser spectroscopy is processed by performing the following operations:

[0111] Acquire real-time data on gas in transformer oil based on laser spectroscopy;

[0112] Cleaning and processing of real-time data on gases in transformer oil based on laser spectroscopy, including:

[0113] Consistency checks were performed on real-time data of gas in transformer oil based on laser spectroscopy.

[0114] Based on the reasonable value range and interrelationship of each variable in the real-time data of gas in transformer oil based on laser spectroscopy, check whether the real-time data of gas in transformer oil based on laser spectroscopy meets the requirements.

[0115] Remove inconsistent data from real-time data on gas in transformer oil based on laser spectroscopy that are outside the normal range, logically unreasonable, or contradictory.

[0116] Invalid and missing values ​​were processed in real-time data of gas in transformer oil based on laser spectroscopy.

[0117] Remove invalid and missing data that are not valuable for the detection of gas in transformer oil based on laser spectroscopy from the real-time data of gas in transformer oil based on laser spectroscopy;

[0118] Real-time data on gas in transformer oil based on laser spectroscopy were identified as valuable for gas detection in transformer oil.

[0119] The real-time data of gas in transformer oil based on laser spectroscopy, which is valuable for gas detection in transformer oil after cleaning, are normalized.

[0120] Eliminate dimensional differences between real-time data on gases in transformer oil based on laser spectroscopy;

[0121] Real-time data on gas in transformer oil based on laser spectroscopy were determined using standardized data.

[0122] Feature extraction was performed on real-time data of gas in transformer oil based on laser spectroscopy, which was standardized by data.

[0123] Extract features that are valuable for detecting gases in transformer oil based on laser spectroscopy;

[0124] Characteristic data of gas in transformer oil based on laser spectroscopy were determined.

[0125] S2: Based on the requirements for gas detection in transformer oil, a fault identification model for gas detection in transformer oil based on laser spectroscopy is constructed. The fault identification model for gas detection in transformer oil based on laser spectroscopy is evaluated and optimized to determine the optimal fault identification model for gas detection in transformer oil.

[0126] In this embodiment, a fault identification model for gas detection in transformer oil based on laser spectroscopy is constructed. The model is then evaluated and optimized by performing the following operations:

[0127] Based on the requirements for gas detection in transformer oil using laser spectroscopy, historical gas concentrations, trends, gas ratios, and gas change rates are collected to determine historical gas data in transformer oil based on laser spectroscopy.

[0128] Historical data on gases in transformer oil based on laser spectroscopy are segmented;

[0129] Determine the gas detection training set and the gas detection test set;

[0130] Based on the requirements for gas detection in transformer oil using laser spectroscopy, a neural network model framework suitable for gas detection in transformer oil using laser spectroscopy is selected.

[0131] Based on the gas detection training set, a selected neural network model framework suitable for gas detection in transformer oil based on laser spectroscopy was trained.

[0132] A fault identification model for gas detection in transformer oil based on laser spectroscopy was determined;

[0133] Based on a gas detection test set, the performance of a fault identification model for gas detection in transformer oil based on laser spectroscopy is evaluated.

[0134] The performance test evaluation results of the fault identification model for gas detection in transformer oil based on laser spectroscopy were determined.

[0135] The performance test and evaluation results of the fault identification model for gas detection in transformer oil based on laser spectroscopy are analyzed.

[0136] A parameter adjustment and optimization scheme for a fault identification model for gas detection in transformer oil based on laser spectroscopy was determined.

[0137] The parameters of the fault identification model for gas detection in transformer oil based on laser spectroscopy are adjusted and optimized according to the parameter adjustment and optimization scheme.

[0138] The optimal fault identification model for gas detection in transformer oil based on laser spectroscopy was determined.

[0139] S3: Based on the optimal fault identification model for gas detection in transformer oil based on laser spectroscopy, predictive analysis is performed on the characteristic data of gas in transformer oil based on laser spectroscopy to determine the fault identification result for gas detection in transformer oil based on laser spectroscopy.

[0140] In this embodiment, predictive analysis is performed on the gas characteristic data in transformer oil based on laser spectroscopy, and the following operations are performed:

[0141] Acquire gas characteristic data in transformer oil based on laser spectroscopy;

[0142] The characteristic data of gas in transformer oil based on laser spectroscopy are input into the optimal fault identification model for gas detection in transformer oil based on laser spectroscopy.

[0143] Based on the optimal fault identification model for gas detection in transformer oil based on laser spectroscopy, predictive analysis is performed on the characteristic data of gas in transformer oil based on laser spectroscopy.

[0144] The results of fault identification based on laser spectroscopy for gas detection in transformer oil were determined.

[0145] S4: Develop an intelligent management and control scheme for gas detection in transformer oil based on laser spectroscopy. Based on this scheme, implement intelligent management and control of gas detection in transformer oil based on laser spectroscopy to ensure long-term stable operation of the transformer.

[0146] In this embodiment, intelligent management and control of gas detection in transformer oil based on laser spectroscopy is implemented by performing the following operations:

[0147] Obtain fault identification results for gas detection in transformer oil based on laser spectroscopy;

[0148] Analysis of fault identification results based on gas detection in transformer oil using laser spectroscopy;

[0149] An intelligent control scheme for gas detection in transformer oil based on laser spectroscopy was determined;

[0150] Intelligent management and control of gas detection in transformer oil based on laser spectroscopy is implemented.

[0151] Specifically, based on the fault identification results of gas detection in transformer oil based on laser spectroscopy, the nature and type of transformer fault are determined, timely early warning is issued, and intelligent control measures are formulated. Transformer faults are repaired in a timely manner, and the transformer is regularly maintained and optimized according to the actual operating conditions to ensure long-term stable operation of the transformer.

[0152] Specifically, real-time data on gases in transformer oil based on laser spectroscopy, which are valuable for detecting gases in transformer oil, were identified, including:

[0153] Real-time monitoring removes invalid data that is of no value to the detection of gases in transformer oil based on laser spectroscopy;

[0154] Extract invalid data that has already been removed;

[0155] Sort the invalid data according to the generation time of each invalid data, and obtain the data set corresponding to the invalid data;

[0156] Retrieve a preset time period; wherein the preset time period ranges from 3 min to 7 min.

[0157] The data set corresponding to the invalid data is divided according to the preset unit time period to obtain multiple invalid data groups;

[0158] The invalid evaluation coefficient for each invalid data group is obtained by utilizing the data generation time interval of the invalid data contained in the invalid data group; wherein, the invalid evaluation coefficient is obtained by the following formula:

[0159] ;

[0160] Where P represents the invalidity evaluation coefficient; n represents the number of invalid data points contained in each invalid data group; T c T represents the duration of a unit of time; i T represents the time of generation corresponding to the i-th invalid data. i-1 Indicates the generation time corresponding to the (i-1)th invalid data; w i w represents the weight value corresponding to the i-th invalid data; i-1 G represents the weight value corresponding to the (i-1)th invalid data point; i G represents the time interval between the generation time of the most recent valid data with the same weight value corresponding to the i-th invalid data; i-1 G represents the time interval between the generation time of the most recent valid data with the same weight value corresponding to the (i-1)th invalid data and the i-th invalid data; p This represents the average of the time intervals between the generation of n invalid data points.

[0161] The invalid evaluation coefficient is compared with a preset invalid evaluation coefficient threshold;

[0162] Extract invalid evaluation coefficients that exceed the invalid evaluation coefficient threshold and use them as target invalid evaluation coefficients;

[0163] The invalid data is filled using the target invalidity evaluation coefficient.

[0164] The technical effects of the above solution are as follows: By real-time monitoring and removal of invalid data that is of no value to the detection of gas in transformer oil based on laser spectroscopy, noise and interference in the data can be significantly reduced, thereby improving the accuracy and reliability of subsequent data analysis. Extracting the removed invalid data and sorting it according to its generation time helps to understand the distribution characteristics and generation patterns of invalid data, providing a foundation for further data processing.

[0165] By dividing invalid data into preset time intervals (3-7 minutes) and calculating the invalid evaluation coefficient for each invalid data group, the impact of invalid data on overall data quality can be assessed more precisely. Comparing the invalid evaluation coefficient with a preset invalid evaluation coefficient threshold intelligently identifies invalid data groups with a significant impact on overall data quality—that is, data groups corresponding to the target invalid evaluation coefficient. By filling invalid data with the target invalid evaluation coefficient, data integrity can be restored to a certain extent, reducing the negative impact on subsequent data analysis. Overall, this technical solution, through a refined invalid data processing strategy, helps improve the real-time performance and accuracy of laser spectroscopy-based gas detection in transformer oil, providing more reliable data support for transformer fault early warning and condition monitoring. Simultaneously, by effectively processing invalid data, this technical solution can also enhance the robustness of the entire detection system, enabling it to maintain more stable and reliable performance in the face of complex and changing actual operating environments.

[0166] In summary, this technical solution significantly improves the data quality and real-time performance of laser spectroscopy-based gas detection in transformer oil through a series of refined data processing steps, providing strong technical support for transformer fault early warning and condition monitoring.

[0167] Specifically, the invalid data is filled using the target invalidity evaluation coefficient, including:

[0168] Extract the invalid data group to which each target invalid evaluation coefficient belongs, and use it as the target invalid data group;

[0169] Extract the generation time corresponding to each invalid data point in the invalid data group to which each invalid evaluation coefficient of the target belongs;

[0170] Based on the generation time of each invalid data, obtain the valid data that comes before the generation time of each invalid data and the valid data that comes after the generation time of each invalid data;

[0171] By combining the valid data that occurred earlier and the valid data that occurred later in time with the time adjacent to the time each invalid data was generated, and the target invalid evaluation coefficient, the invalid data compensation coefficient corresponding to the target invalid data group is obtained.

[0172] The invalid data compensation coefficient corresponding to the target invalid data group is obtained by the following formula:

[0173] ;

[0174] Where B represents the invalid data compensation coefficient corresponding to the target invalid data group; X 01 and X 02These represent the data values ​​of the valid data occurring before and after the time of each invalid data occurrence; X p P represents the average value of valid data; P represents the invalid evaluation coefficient; P0 represents the preset invalid evaluation coefficient threshold; X z This represents the intermediate value of the valid data.

[0175] The replacement data corresponding to each invalid data in the target invalid data group is obtained using the invalid data compensation coefficient corresponding to the target invalid data group;

[0176] Replace each invalid data in the target invalid data group with the replacement data corresponding to each invalid data in the target invalid data group.

[0177] The technical effect of the above solution is as follows: by extracting the target invalid data group and replacing the invalid data within it, the missing values ​​in the dataset can be effectively reduced, thereby improving data integrity. This is crucial for subsequent data analysis and model training, as missing data may lead to inaccurate results or degraded model performance. Calculating the invalid data compensation coefficient using adjacent valid data (i.e., valid data from earlier and later times) and the target invalid evaluation coefficient allows for a more accurate estimation of the reasonable values ​​of invalid data. This method considers the temporal sequence of the data and the correlation between adjacent data, helping to generate replacement data that more closely approximates the true values.

[0178] This technical solution achieves intelligent identification and processing of invalid data by calculating invalid evaluation coefficients and invalid data compensation coefficients. Compared to manually deleting or simply filling in invalid data, this method is more efficient and retains more data information and structure. In applications such as gas detection in transformer oil based on laser spectroscopy, the accuracy and completeness of data are crucial to the predictive accuracy of the model. By filling in invalid data, noise and missing values ​​in the data can be reduced, thereby improving the predictive performance of the model.

[0179] This technical solution enhances the robustness of the entire system through a refined data processing strategy. Even when encountering a large amount of invalid data during actual operation, the system can maintain data stability and reliability through intelligent identification and filling strategies. Furthermore, this technical solution is not only applicable to gas detection in transformer oil based on laser spectroscopy, but can also be extended to other scenarios requiring the processing of invalid data. By adjusting parameters and formulas, it can be adapted to different datasets and application requirements.

[0180] In summary, this technical solution significantly improves data integrity, quality, and processing efficiency by utilizing target invalid evaluation coefficients and invalid data compensation coefficients to fill in invalid data, providing more reliable data support for applications such as laser spectroscopy-based gas detection in transformer oil. Furthermore, this solution enhances system robustness and provides flexible data processing options.

[0181] Specifically, the replacement data corresponding to each invalid data in the target invalid data group is obtained using the invalid data compensation coefficient corresponding to the target invalid data group, including:

[0182] Retrieve the invalid data compensation coefficients corresponding to the target invalid data group;

[0183] Retrieve valid data from each invalid data group in the target invalid data group, including valid data with the earliest and latest times that are adjacent to each invalid data point.

[0184] By utilizing the valid data that occurs earlier and the valid data that occurs later in the time adjacent to the time of each invalid data in the target invalid data group, the data baseline value corresponding to each invalid data in the target invalid data group is obtained;

[0185] The data baseline value is obtained using the following formula:

[0186] ;

[0187] Among them, X k This represents the baseline value corresponding to each invalid data point; X 01 and X 02 These represent the data values ​​of the valid data occurring earlier and later in the time frame adjacent to the time each invalid data point was generated; P0 represents the preset invalid evaluation coefficient threshold; P z This represents the median value of the invalid evaluation coefficients corresponding to all invalid data groups;

[0188] The replacement data for each invalid data is obtained by combining the invalid data compensation coefficient corresponding to the target invalid data group with the data baseline value corresponding to each invalid data in the target invalid data group;

[0189] The replacement data corresponding to each invalid data is obtained by the following formula:

[0190] ;

[0191] Among them, X t This represents the replacement data for each invalid data; X k This represents the baseline value corresponding to each invalid data point; X 01 and X 02represents the data values ​​of the valid data that occurred earlier and later in the time frame adjacent to the time each invalid data was generated; B represents the invalid data compensation coefficient corresponding to the target invalid data group.

[0192] The technical effect of the above solution is as follows: By combining adjacent valid data (valid data earlier and later in time) with a preset invalid evaluation coefficient threshold and the median invalid evaluation coefficient of all invalid data groups, a data baseline value corresponding to each invalid data point is calculated. This method not only considers the temporal sequence of the data but also incorporates the overall evaluation of invalid data, making the generated data baseline value closer to the true value, thereby improving data accuracy. Using the data baseline value and invalid data compensation coefficient to generate replacement data allows for more flexible handling of invalid data. This method not only considers the characteristics of the invalid data itself but also incorporates information from adjacent valid data, making the generated replacement data more consistent with the overall distribution and trend of the data.

[0193] This technique helps maintain data consistency by intelligently identifying and filling in invalid data. The replaced data is numerically and trend-wise closer to adjacent valid data, thus reducing abrupt changes and outliers and improving overall data quality. In data analysis and machine learning models, the accuracy and completeness of data are crucial to model performance. By replacing invalid data, noise and missing values ​​can be reduced, thereby improving model training effectiveness and prediction accuracy.

[0194] This technical solution achieves rapid identification and replacement of invalid data through an automated data processing workflow. Compared to manual methods of handling invalid data, this approach is more efficient and capable of processing large datasets. Through refined data processing strategies, this solution enhances the robustness of the entire system. Even when encountering large amounts of invalid data in actual operation, the system can maintain data stability and reliability through intelligent identification and replacement strategies.

[0195] In summary, this technical solution generates replacement data by calculating baseline data values ​​and invalid data compensation coefficients, significantly improving data accuracy, consistency, and overall quality. Simultaneously, this solution also enhances model performance, data processing efficiency, and system robustness, providing more reliable data support for data analysis and machine learning applications.

[0196] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0197] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for detecting gas in transformer oil based on laser spectroscopy, characterized in that, Includes the following steps: S1: Collect real-time data of gas in transformer oil based on laser spectroscopy, process the real-time data of gas in transformer oil based on laser spectroscopy, and determine the characteristic data of gas in transformer oil based on laser spectroscopy. S2: Based on the requirements for gas detection in transformer oil based on laser spectroscopy, a fault identification model for gas detection in transformer oil based on laser spectroscopy is constructed. The fault identification model for gas detection in transformer oil based on laser spectroscopy is evaluated and optimized to determine the optimal fault identification model for gas detection in transformer oil based on laser spectroscopy. S3: Based on the optimal fault identification model for gas detection in transformer oil based on laser spectroscopy, predictive analysis is performed on the characteristic data of gas in transformer oil based on laser spectroscopy to determine the fault identification result for gas detection in transformer oil based on laser spectroscopy. S4: Develop an intelligent management and control scheme for gas detection in transformer oil based on laser spectroscopy. Based on the intelligent management and control scheme for gas detection in transformer oil based on laser spectroscopy, implement intelligent management and control for gas detection in transformer oil based on laser spectroscopy to ensure long-term stable operation of the transformer. In step S1, the real-time data of gas in transformer oil based on laser spectroscopy is cleaned to determine the real-time data of gas in transformer oil based on laser spectroscopy that is valuable for the detection of gas in transformer oil based on laser spectroscopy, including: Real-time monitoring removes invalid data that is of no value to gas detection in transformer oil based on laser spectroscopy; extracts the removed invalid data; sorts the invalid data according to the generation time of each invalid data to obtain the data set corresponding to the invalid data; Retrieve a preset time period; wherein the preset time period ranges from 3 min to 7 min; divide the data set corresponding to the invalid data according to the preset time period to obtain multiple invalid data groups; The invalid evaluation coefficient for each invalid data group is obtained by utilizing the data generation time interval of the invalid data contained in the invalid data group; wherein, the invalid evaluation coefficient is obtained by the following formula: ； Where P represents the invalidity evaluation coefficient; n represents the number of invalid data points contained in each invalid data group; T c T represents the duration of a unit of time; i T represents the time of generation corresponding to the i-th invalid data. i-1 Indicates the generation time corresponding to the (i-1)th invalid data; w i w represents the weight value corresponding to the i-th invalid data; i-1 G represents the weight value corresponding to the (i-1)th invalid data point; i G represents the time interval between the generation time of the most recent valid data with the same weight value corresponding to the i-th invalid data; i-1 G represents the time interval between the generation time of the most recent valid data with the same weight value corresponding to the (i-1)th invalid data and the i-th invalid data; p This represents the average of the time intervals between the generation of n invalid data points. The invalid evaluation coefficient is compared with a preset invalid evaluation coefficient threshold; invalid evaluation coefficients that exceed the invalid evaluation coefficient threshold are extracted as target invalid evaluation coefficients. Filling in invalid data using the target invalidity evaluation coefficient includes: Extract the invalid data group to which each invalid evaluation coefficient belongs, and use it as the invalid data group for each target; extract the generation time corresponding to each invalid data in the invalid data group to which each invalid evaluation coefficient belongs; Based on the generation time of each invalid data, obtain the valid data that comes before the generation time of each invalid data and the valid data that comes after the generation time of each invalid data; By combining the valid data that occurred earlier and the valid data that occurred later in time with the time adjacent to the time each invalid data was generated, and the target invalid evaluation coefficient, the invalid data compensation coefficient corresponding to the target invalid data group is obtained. The replacement data corresponding to each invalid data in the target invalid data group is obtained using the invalid data compensation coefficient corresponding to the target invalid data group; the replacement data corresponding to each invalid data in the target invalid data group is used to replace each invalid data in the target invalid data group.

2. The method for detecting gas in transformer oil based on laser spectroscopy according to claim 1, characterized in that, In step S1, real-time data on gas in transformer oil based on laser spectroscopy is collected, and the following operations are performed: The transformer oil sample is determined by taking a sample directly from the oil sampling port on the transformer. A constant-temperature vacuum degassing technique was used to efficiently degas transformer oil samples. The gas dissolved in the transformer oil sample is released, while preventing the transformer oil sample from contacting the air. The degassed gas is then introduced into the photoacoustic cell. A laser of a specific wavelength is emitted to irradiate the gas in the photoacoustic cell. After the gas molecules absorb the laser energy, their temperature rises and they release heat energy, which in turn generates pressure waves to determine the gas concentration. Obtain gas concentration, trends, gas ratios, and gas change rates; Real-time data on gas in transformer oil based on laser spectroscopy were determined.

3. The method for detecting gas in transformer oil based on laser spectroscopy according to claim 2, characterized in that, In step S1, the real-time data on gas in transformer oil based on laser spectroscopy is processed, and the following operations are performed: Acquire real-time data on gas in transformer oil based on laser spectroscopy; Cleaning and processing of real-time data on gases in transformer oil based on laser spectroscopy, including: Consistency checks were performed on real-time data of gas in transformer oil based on laser spectroscopy. Based on the reasonable value range and interrelationship of each variable in the real-time data of gas in transformer oil based on laser spectroscopy, check whether the real-time data of gas in transformer oil based on laser spectroscopy meets the requirements. Remove inconsistent data from real-time data on gas in transformer oil based on laser spectroscopy that are outside the normal range, logically unreasonable, or contradictory. Invalid and missing values ​​were processed in real-time data of gas in transformer oil based on laser spectroscopy. Remove invalid and missing data that are not valuable for the detection of gas in transformer oil based on laser spectroscopy from the real-time data of gas in transformer oil based on laser spectroscopy; Real-time data on gas in transformer oil based on laser spectroscopy were identified as valuable for gas detection in transformer oil.

4. The method for detecting gas in transformer oil based on laser spectroscopy according to claim 3, characterized in that, The invalid data compensation coefficient corresponding to the target invalid data group is obtained by the following formula: ； Where B represents the invalid data compensation coefficient corresponding to the target invalid data group; X 01 and X 02 These represent the data values ​​of the valid data occurring before and after the time of each invalid data occurrence; X p P represents the average value of valid data; P represents the invalid evaluation coefficient; P0 represents the preset invalid evaluation coefficient threshold; X z This represents the intermediate value of the valid data.

5. The method for detecting gas in transformer oil based on laser spectroscopy according to claim 4, characterized in that, Using the invalid data compensation coefficient corresponding to the target invalid data group, the replacement data corresponding to each invalid data in the target invalid data group is obtained, including: Retrieve the invalid data compensation coefficients corresponding to the target invalid data group; Retrieve valid data from each invalid data group in the target invalid data group, including valid data with the earliest and latest times that are adjacent to each invalid data point. By utilizing the valid data that occurs earlier and the valid data that occurs later in the time adjacent to the time of each invalid data in the target invalid data group, the data baseline value corresponding to each invalid data in the target invalid data group is obtained; The data baseline value is obtained using the following formula: ； Among them, X k This represents the baseline value corresponding to each invalid data point; X 01 and X 02 These represent the data values ​​of the valid data occurring earlier and later in the time frame adjacent to the time each invalid data point was generated; P0 represents the preset invalid evaluation coefficient threshold; P z This represents the median value of the invalid evaluation coefficients corresponding to all invalid data groups; The replacement data for each invalid data is obtained by combining the invalid data compensation coefficient corresponding to the target invalid data group with the data baseline value corresponding to each invalid data in the target invalid data group; The replacement data corresponding to each invalid data is obtained by the following formula: ； Among them, X t This represents the replacement data for each invalid data; X k This represents the baseline value corresponding to each invalid data point; X 01 and X 02 represents the data values ​​of the valid data that occurred earlier and later in the time frame adjacent to the time each invalid data was generated; B represents the invalid data compensation coefficient corresponding to the target invalid data group.

6. The method for detecting gas in transformer oil based on laser spectroscopy according to claim 5, characterized in that, In step S1, the real-time data on gas in transformer oil based on laser spectroscopy is processed, and the following operations are also performed: To acquire real-time data on gas in transformer oil based on laser spectroscopy, which is valuable for gas detection in transformer oil. The real-time data of gas in transformer oil based on laser spectroscopy, which is valuable for gas detection in transformer oil after cleaning, are normalized. Eliminate dimensional differences between real-time data on gases in transformer oil based on laser spectroscopy; Real-time data on gas in transformer oil based on laser spectroscopy were determined using standardized data. Feature extraction was performed on real-time data of gas in transformer oil based on laser spectroscopy, which was standardized by data. Extract features that are valuable for detecting gases in transformer oil based on laser spectroscopy; Characteristic data of gas in transformer oil based on laser spectroscopy were determined.

7. The method for detecting gas in transformer oil based on laser spectroscopy according to claim 6, characterized in that, In step S2, a fault identification model for detecting gas in transformer oil based on laser spectroscopy is constructed. The model is then evaluated and optimized, and the following operations are performed: Based on the requirements for gas detection in transformer oil using laser spectroscopy, historical gas concentrations, trends, gas ratios, and gas change rates are collected to determine historical gas data in transformer oil based on laser spectroscopy. Historical data on gases in transformer oil based on laser spectroscopy are segmented; Determine the gas detection training set and the gas detection test set; Based on the requirements for gas detection in transformer oil using laser spectroscopy, a neural network model framework suitable for gas detection in transformer oil using laser spectroscopy is selected. Based on the gas detection training set, a selected neural network model framework suitable for gas detection in transformer oil based on laser spectroscopy was trained. A fault identification model for gas detection in transformer oil based on laser spectroscopy was determined; Based on a gas detection test set, the performance of a fault identification model for gas detection in transformer oil based on laser spectroscopy is evaluated. The performance test evaluation results of the fault identification model for gas detection in transformer oil based on laser spectroscopy were determined. The performance test and evaluation results of the fault identification model for gas detection in transformer oil based on laser spectroscopy are analyzed. A parameter adjustment and optimization scheme for a fault identification model for gas detection in transformer oil based on laser spectroscopy was determined. The parameters of the fault identification model for gas detection in transformer oil based on laser spectroscopy are adjusted and optimized according to the parameter adjustment and optimization scheme. The optimal fault identification model for gas detection in transformer oil based on laser spectroscopy was determined.

8. The method for detecting gas in transformer oil based on laser spectroscopy according to claim 7, characterized in that, In step S3, predictive analysis is performed on the gas characteristic data in transformer oil based on laser spectroscopy, and the following operations are performed: Acquire gas characteristic data in transformer oil based on laser spectroscopy; The characteristic data of gas in transformer oil based on laser spectroscopy are input into the optimal fault identification model for gas detection in transformer oil based on laser spectroscopy. Based on the optimal fault identification model for gas detection in transformer oil based on laser spectroscopy, predictive analysis is performed on the characteristic data of gas in transformer oil based on laser spectroscopy. The results of fault identification based on laser spectroscopy for gas detection in transformer oil were determined.

9. The method for detecting gas in transformer oil based on laser spectroscopy according to claim 8, characterized in that, In step S4, intelligent control is implemented for the detection of gas in transformer oil based on laser spectroscopy, and the following operations are performed: Obtain fault identification results for gas detection in transformer oil based on laser spectroscopy; Analysis of fault identification results based on gas detection in transformer oil using laser spectroscopy; An intelligent control scheme for gas detection in transformer oil based on laser spectroscopy was determined; Intelligent management and control of gas detection in transformer oil based on laser spectroscopy is implemented. Specifically, based on the fault identification results of gas detection in transformer oil based on laser spectroscopy, the nature and type of transformer fault are determined, timely early warning is issued, and intelligent control measures are formulated. Transformer faults are repaired in a timely manner, and the transformer is regularly maintained and optimized according to the actual operating conditions to ensure long-term stable operation of the transformer.

Images