Method for detecting hydrogen in transformer oil based on big data analysis and wavelet analysis

Abstract

The invention discloses a method for detecting hydrogen in transformer oil based on big data analysis and wavelet analysis, and belongs to the field of transformer oil detection. The hydrogen detection method comprises the following steps: collecting and preprocessing gas data in transformer oil; smoothing the preprocessed data by adopting a wavelet analysis method; feature extraction is carried out on the smoothed data, and key parameters related to hydrogen concentration changes are recognized; training a support vector machine model by using the extracted features to establish a relation model between the hydrogen concentration and other variables; the smoothed data are input into a trained support vector machine model, the hydrogen concentration is predicted in real time, and fault types such as partial discharge, overheating and arc discharge can be accurately identified by combining wavelet analysis and monitoring a sudden change point in a hydrogen concentration signal; meanwhile, the incidence relation among the hydrogen concentration, the temperature, the load rate and other operation parameters is established through big data analysis, and conversion from passive response to active prevention is achieved.

Description

Technical Field

[0001] The present invention belongs to the field of transformer oil detection, and in particular relates to a method for detecting hydrogen in transformer oil based on big data analysis and wavelet analysis. Background Art

[0002] During the use of large transformers, various factors such as local heat release and aging will cause the insulating oil and the organic matter inside to produce hydrogen and short-chain hydrocarbons. This gas (mainly hydrogen) plays a very important role in the early fault diagnosis of transformers and is called "fault gas". Existing hydrogen detection systems often only issue warnings when the hydrogen concentration exceeds the standard. They lack the ability to predict and prevent in advance, and cannot take effective maintenance measures in time, thereby increasing the risk of transformer failure. Therefore, it is very necessary to design an efficient online detection system for hydrogen concentration in transformer oil.

[0003] Experts have conducted special research on hydrogen detection in transformer oil for a long time. Hydrogen content is reduced by adsorbing hydrogen in transformer oil with zinc oxide, titanium dioxide, carbon nanotubes and nanoparticles. Infrared spectroscopy measurement is used to replace dissolved gas analysis (DGA) to diagnose power transformer insulation oil faults. Fourier transform infrared (FTIR) is used to simulate the behavior of oil samples and conduct preliminary verification of hydrogen detection. Three Duval triangles and an empirically evaluated triangular fuzzy function are combined to diagnose the initial fault of the transformer. Some researchers use the intelligent software ANFIS Expert system and dissolved gas analysis methods to diagnose transformer oil faults and quality assessment. However, although the existing gas adsorption and sensing technology can reduce the hydrogen content and improve the thermal conductivity, the accuracy and stability of gas detection in practical applications are still limited. Secondly, the traditional dissolved gas analysis method has the problems of complexity and high cost, and the emerging infrared spectroscopy measurement and other alternative methods, although promising, have not yet completely solved the problem of accuracy in fault diagnosis. Summary of the invention

[0004] In view of the deficiencies in the prior art, the purpose of the present invention is to provide a method for detecting hydrogen in transformer oil based on big data analysis and wavelet analysis, which solves the problems in the prior art.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] The method for detecting hydrogen in transformer oil based on big data analysis and wavelet analysis includes the following steps:

[0007] Collect gas data in transformer oil and perform pre-processing;

[0008] Wavelet analysis is used to smooth the preprocessed data;

[0009] By extracting features from the smoothed data, key parameters related to changes in hydrogen concentration are identified;

[0010] The extracted features were used to train a support vector machine model to model the relationship between hydrogen concentration and other variables;

[0011] The smoothed data is input into the trained support vector machine model to predict the hydrogen concentration in real time, and combined with wavelet analysis, the mutation points in the hydrogen concentration signal are monitored.

[0012] Furthermore, data preprocessing includes: collecting and analyzing the historical operation data of the transformer, studying the influence of operating parameters such as temperature and load rate on the gas in the oil, comparing the changing trend of gas concentration under different working conditions, and comprehensively considering the fault characteristics such as partial discharge, overheating and arc discharge inside the transformer, and determining the influencing factors of dissolved gas in transformer oil as follows:

[0013] [x1,x2,x3,x4,…x 40 ,…x 49 ,y];

[0014] Among them, [x1,x2,x3,x4,…x 40 ] is the historical data value of gas in transformer oil, [x 41 ,…x 49 ] is the current time value of H2, CH4, C2H6, C2H4 and C2H2 concentrations, as well as the main transformer active power, top oil temperature, oil level of the oil pillow and left temperature; y is the gas value corresponding to the predicted time.

[0015] Furthermore, the process of data smoothing is:

[0016] 1) Use Mallat algorithm to solve [x1,x2,x3,x4,…x 40 ] is decomposed into three layers of high frequency and low frequency:

[0017]

[0018] Where H is a low-pass filter; G is a high-pass filter; J is the maximum scale resolution; c j The resolution is 2 -j The low-frequency component of the signal; d j The resolution is 2 -j The high frequency component of the signal;

[0019] 2) Reconstruct high and low frequency signals;

[0020] c j =H * c j+1 +G* d j+1 ,j=1,0

[0021] In the formula, H * is the dual operator of H; G * is the dual operator of G; reconstruct c1 and d1 respectively to obtain the reconstructed low-frequency signal and high-frequency signal; decompose the obtained low-frequency signal twice to obtain the second low-frequency signal and the second high-frequency signal, as well as the third low-frequency signal and the third high-frequency signal; obtain four wavelet sample sets:

[0022]

[0023] Furthermore, the feature extraction process includes:

[0024] 1) For four wavelet sample sets Perform feature analysis to extract time domain features, frequency domain features, and statistical features;

[0025] 2) Combine [x 41 …x 49 ] to construct the eigenvector matrix.

[0026] Furthermore, the model training process is:

[0027] 1) Divide the feature vector matrix into training set and test set in a ratio of 7:3;

[0028] 2) Using support vector machine as the basic model, using radial basis kernel function, and optimizing kernel parameter σ and penalty factor C through cross validation;

[0029] 3) Use the training set data to train the model, and use the root mean square error and determination coefficient as model evaluation indicators.

[0030] Furthermore, the process of hydrogen concentration prediction and mutation point detection is as follows:

[0031] After the real-time collected data has gone through the same preprocessing and feature extraction process, it is input into the trained SVM model to obtain the predicted hydrogen concentration value; the predicted hydrogen concentration sequence is decomposed using cubic spline wavelet to detect mutation points.

[0032] The above-mentioned hydrogen detection method in transformer oil based on big data analysis and wavelet analysis is applied in transformer fault detection.

[0033] Transformer oil hydrogen detection system based on big data analysis and wavelet analysis, including:

[0034] Data acquisition module: collects gas data in transformer oil and performs preprocessing;

[0035] Data processing module: Use wavelet analysis to smooth the pre-processed data;

[0036] Feature extraction module: By extracting features from the smoothed data, key parameters related to changes in hydrogen concentration are identified;

[0037] Model training module: Use the extracted features to train the support vector machine model to establish a relationship model between hydrogen concentration and other variables;

[0038] And, monitoring module: input the smoothed data into the trained support vector machine model to predict the hydrogen concentration in real time, and combine with wavelet analysis to monitor the mutation points in the hydrogen concentration signal.

[0039] A computer storage medium stores a readable program, which can execute the above-mentioned method for detecting hydrogen in transformer oil based on big data analysis and wavelet analysis when the program is running.

[0040] An electronic device, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other through the communication bus;

[0041] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the above-mentioned method for detecting hydrogen in transformer oil based on big data analysis and wavelet analysis.

[0042] Beneficial effects of the present invention:

[0043] 1. The present invention combines big data analysis and wavelet analysis to design an efficient transformer oil hydrogen detection method. Through wavelet analysis, effective data smoothing and accurate positioning of mutation points are achieved, and fault types such as partial discharge, overheating and arc discharge can be accurately identified; at the same time, the correlation between hydrogen concentration and operating parameters such as temperature and load rate is established by big data analysis, realizing the transition from passive response to active prevention. This method improves the detection accuracy of hydrogen concentration from the traditional ppm level to a more accurate level, significantly reduces the false alarm and missed alarm rates, realizes real-time monitoring and evaluation of the transformer operating status, and provides a stronger guarantee for the safe operation of the transformer.

[0044] 2. Online monitoring of hydrogen concentration can effectively identify the type and severity of transformer faults, which has important practical application value and helps maintenance personnel take timely measures to extend the service life of the transformer.

[0045] 3. The present invention combines big data analysis and wavelet analysis to design an efficient transformer oil hydrogen detection method. Through wavelet analysis, effective smoothing of data and accurate positioning of mutation points are achieved, and fault types such as partial discharge, overheating and arc discharge can be accurately identified; at the same time, the correlation between hydrogen concentration and operating parameters such as temperature and load rate is established by big data analysis, realizing the transition from passive response to active prevention. This method improves the detection accuracy of hydrogen concentration from the traditional ppm level to a more accurate level, significantly reducing the false alarm and missed alarm rate. In addition, the present invention uses wavelet analysis to standardize and smooth the data, and combines machine learning models for automated analysis, avoiding the need to use expensive analytical equipment such as gas chromatography in traditional methods, and realizing data analysis and fault diagnosis through software algorithms, reducing hardware investment and manual judgment costs; in addition, the system realizes online monitoring and early warning functions, avoiding frequent offline sampling and laboratory analysis, and significantly reducing the complexity and cost of detection and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0047] Figure 1 is the historical data of hydrogen concentration in transformer oil;

[0048] Figure 2 The relationship between hydrogen concentration in transformer oil and insulation aging state;

[0049] Figure 3 Detection of hydrogen concentration mutation point. DETAILED DESCRIPTION

[0050] 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.

[0051] Example 1

[0052] The method for detecting hydrogen in transformer oil based on big data analysis and wavelet analysis includes the following steps:

[0053] S1, collect historical data of gas in transformer oil and perform preprocessing;

[0054] Data preprocessing includes: collecting and analyzing the historical operation data of the transformer, studying the influence of operating parameters such as temperature and load rate on the gas in the oil, comparing the changing trend of gas concentration under different working conditions, and comprehensively considering the fault characteristics such as partial discharge, overheating and arc discharge inside the transformer, and determining the influencing factors of dissolved gas in transformer oil as follows:

[0055] [x1,x2,x3,x4,…x 40 ,…x 49 ,y];

[0056] Among them, [x1,x2,x3,x4,…x 40 ] is the historical data value of gas in transformer oil, [x 41 ,…x 49 ] is the current time value of H2, CH4, C2H6, C2H4 and C2H2 concentrations, as well as the main transformer active power, top oil temperature, oil level of the oil pillow and left temperature; y is the gas value corresponding to the predicted time.

[0057] S2, smoothing the preprocessed data using wavelet analysis;

[0058] The specific processing process is as follows:

[0059] 1) Use Mallat algorithm to solve [x1,x2,x3,x4,…x 40 ] is decomposed into three layers of high frequency and low frequency:

[0060]

[0061] Where H is a low-pass filter; G is a high-pass filter; J is the maximum scale resolution; c j The resolution is 2 -j The low-frequency component of the signal; d j The resolution is 2 -j The high frequency component of the signal;

[0062] 2) Reconstruct high and low frequency signals.

[0063] c j =H * c j+1 +G * d j+1 ,j=1,0

[0064] In the formula, H * is the dual operator of H; G *is the dual operator of G; reconstruct c1 and d1 respectively to obtain the reconstructed low-frequency signal and high-frequency signal; decompose the obtained low-frequency signal twice to obtain the second low-frequency signal and the second high-frequency signal, as well as the third low-frequency signal and the third high-frequency signal; obtain four wavelet sample sets:

[0065]

[0066] S3, identifying key parameters related to the change of hydrogen concentration by extracting features from the smoothed data;

[0067] The feature extraction process includes:

[0068] 1) For four wavelet sample sets Perform feature analysis to extract time domain features (mean, variance, peak), frequency domain features (energy, entropy) and statistical features (skewness, kurtosis);

[0069] 2) Combine [x 41 …x 49 ] and operating parameters (main transformer active power, top oil temperature, conservator oil level, left side temperature) are used to construct the eigenvector matrix.

[0070] S4, using the extracted features to train a support vector machine (SVM) model to establish a relationship model between hydrogen concentration and other variables;

[0071] The model training process is:

[0072] 1) Divide the feature vector matrix into training set and test set in a ratio of 7:3;

[0073] 2) Support vector machine (SVM) is used as the basic model, radial basis kernel function (RBF) is used, and kernel parameter σ and penalty factor C are optimized through cross-validation;

[0074] 3) Use the training set data to train the model and use the root mean square error (RMSE) and determination coefficient (R 2 ) as the model evaluation indicator.

[0075] S5, inputting the smoothed data into the trained model to predict the hydrogen concentration in real time, and combining with wavelet analysis to monitor the mutation points in the hydrogen concentration signal;

[0076] The process of hydrogen concentration prediction and mutation point detection is as follows:

[0077] 1) After the real-time collected data has gone through the same preprocessing and feature extraction process, it is input into the trained SVM model to obtain the predicted hydrogen concentration value;

[0078] 2) Use cubic spline wavelet decomposition to detect mutation points in the predicted hydrogen concentration sequence:

[0079] 2.1) Find the modulus maximum point at three scales j = 1, 2, and 3;

[0080] 2.2) If the modulus maxima points on adjacent scales are close in position, have the same sign and similar amplitude, they are determined to be mutation points;

[0081] 2.3) Determine the fault type (partial discharge, overheating fault or arc discharge) based on the amplitude and duration of the mutation point.

[0082] Example 2

[0083] In this embodiment, the application of hydrogen concentration detection in transformer fault detection is analyzed. Through the detection methods of S1 to S5 above, real-time monitoring of hydrogen concentration in transformer oil and fault warning can be achieved. This embodiment will demonstrate the detection effect of this method in practical application, verify the corresponding relationship between hydrogen concentration change and transformer fault type, and the influence of operating parameters such as temperature and load on hydrogen concentration, thereby proving the effectiveness and practical value of this detection method in transformer fault diagnosis.

[0084] The experimental data in this embodiment is derived from the regular monitoring of the hydrogen concentration in the transformer oil. The patented data uses a high-precision gas detection instrument that can accurately measure the hydrogen concentration in the transformer oil and record relevant data. At the same time, an insulation paper polymerization tester is used to evaluate the aging degree of the insulation material. The fault diagnosis equipment includes a wavelet analysis instrument to detect sudden changes in hydrogen concentration and analyze the causes. All equipment is calibrated to ensure the accuracy and reliability of the data.

[0085] Figure 1 This is the historical data of hydrogen concentration in transformer oil. It can be observed that from January 2023 to June 2023, the hydrogen concentration in transformer oil gradually increased over time, from 15ppm to 35ppm. This indicates that there may be aging or fault problems inside the transformer, which requires further monitoring and analysis.

[0086] Figure 2 By comparing the relationship between hydrogen concentration and the degree of polymerization of insulation paper, the relationship between hydrogen concentration and the aging state of insulation paper is revealed. The data show that as the hydrogen concentration increases, the degree of polymerization of insulation paper gradually decreases, and the insulation aging state changes from mild aging to moderate aging. It is confirmed that hydrogen concentration is an important indicator for judging transformer insulation aging.

[0087] Figure 3It shows that sudden changes in hydrogen concentration were detected in April, May and June 2023, corresponding to partial discharge, overheating fault and arc discharge; this shows that wavelet analysis can effectively identify the type and time of transformer faults.

[0088] The fault diagnosis results of online monitoring of hydrogen concentration are shown in Table 1:

[0089] Table 1 Hydrogen concentration online monitoring and fault diagnosis results

[0090]

[0091] The results in Table 1 show that when the hydrogen concentration is low, the transformer is in a fault-free state; as the hydrogen concentration increases, the type of fault becomes increasingly obvious, and the severity of the fault also increases. This shows that online monitoring of hydrogen concentration plays an important role in transformer fault diagnosis, and can detect and solve potential problems in a timely manner.

[0092] In order to more comprehensively analyze the change of hydrogen concentration and its influence on transformer performance, in this embodiment, the experimental data is expanded to add records of temperature and transformer load rate, as shown in Table 2:

[0093] Table 2 Relationship between hydrogen concentration, temperature and load rate

[0094]

[0095] The data in Table 2 show that the concentration of hydrogen shows a clear upward trend as the temperature increases. For example, when the temperature rises from 45°C to 65°C, the concentration of hydrogen rises from 15ppm to 35ppm. This phenomenon shows that the increase in temperature may accelerate the aging and decomposition of the insulation material inside the transformer, leading to the release of hydrogen.

[0096] In addition, the increase in transformer load rate is accompanied by an increase in hydrogen concentration, especially when the load rate increases from 60% to 85%, the hydrogen concentration increases significantly. This reflects that high load conditions may aggravate the thermal and electrical stresses inside the transformer, thereby accelerating insulation aging and causing failures.

[0097] The method of the present invention may be implemented in hardware, firmware, or as software or computer code that may be stored in a recording medium (such as a CDROM, RAM, floppy disk, hard disk, or magneto-optical disk), or as computer code that is originally stored in a remote recording medium or a non-temporary machine-readable medium downloaded over a network and will be stored in a local recording medium, so that the method described herein may be stored in such software processing on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA). It is understood that a computer, processor, microprocessor controller, or programmable hardware includes a storage component (e.g., RAM, ROM, flash memory, etc.) that can store or receive software or computer code, and when the software or computer code is accessed and executed by a computer, processor, or hardware, the method described herein is implemented. In addition, when a general-purpose computer accesses the code for implementing the method shown herein, the execution of the code converts the general-purpose computer into a dedicated computer for executing the method shown herein.

[0098] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.

Claims

1. A method for detecting hydrogen in transformer oil based on big data analysis and wavelet analysis, characterized in that: The following steps are involved: Collect gas data in transformer oil and perform pre-processing; Wavelet analysis is used to smooth the preprocessed data; By extracting features from the smoothed data, key parameters related to changes in hydrogen concentration are identified; The extracted features were used to train a support vector machine model to model the relationship between hydrogen concentration and other variables; The smoothed data is input into the trained support vector machine model to predict the hydrogen concentration in real time, and combined with wavelet analysis, the mutation points in the hydrogen concentration signal are monitored.

2. The method for detecting hydrogen in transformer oil based on big data analysis and wavelet analysis according to claim 1, characterized in that: Data preprocessing includes: collecting and analyzing the historical operation data of the transformer, studying the influence of operating parameters such as temperature and load rate on the gas in the oil, comparing the changing trend of gas concentration under different working conditions, and comprehensively considering the fault characteristics such as partial discharge, overheating and arc discharge inside the transformer, and determining the influencing factors of dissolved gas in transformer oil as follows: [x1,x2,x3,x4,…x 40 ,…x 49 ,y]; Among them, [x1,x2,x3,x4,…x 40 ] is the historical data value of gas in transformer oil, [x 41 ,…x 49 ] is the current time value of H2, CH4, C2H6, C2H4 and C2H2 concentrations, as well as the main transformer active power, top oil temperature, oil level of the oil pillow and left temperature; y is the gas value corresponding to the predicted time.

3. The method for detecting hydrogen in transformer oil based on big data analysis and wavelet analysis according to claim 2 is characterized in that: The process of data smoothing is: 1) Use Mallat algorithm to solve [x1,x2,x3,x4,…x 40 ] is decomposed into three layers of high frequency and low frequency: Where H is a low-pass filter; G is a high-pass filter; J is the maximum scale resolution; c j The resolution is 2 -j The low-frequency component of the signal; d j The resolution is 2 -j The high frequency component of the signal; 2) Reconstruct high and low frequency signals; c j =H * c j+1 +G * d j+1 ,j=1,0 In the formula, H * is the dual operator of H; G * is the dual operator of G; reconstruct c1 and d1 respectively to obtain the reconstructed low-frequency signal and high-frequency signal; decompose the obtained low-frequency signal twice to obtain the second low-frequency signal and the second high-frequency signal, as well as the third low-frequency signal and the third high-frequency signal; obtain four wavelet sample sets:

4. The method for detecting hydrogen in transformer oil based on big data analysis and wavelet analysis according to claim 3 is characterized in that: The feature extraction process includes: 1) For four wavelet sample sets Perform feature analysis to extract time domain features, frequency domain features, and statistical features; 2) Combine [x 41 …x 49 ] to construct the eigenvector matrix.

5. The method for detecting hydrogen in transformer oil based on big data analysis and wavelet analysis according to claim 4 is characterized in that: The model training process is: 1) Divide the feature vector matrix into training set and test set in a ratio of 7:3; 2) Using support vector machine as the basic model, using radial basis kernel function, and optimizing kernel parameter σ and penalty factor C through cross validation; 3) Use the training set data to train the model, and use the root mean square error and determination coefficient as model evaluation indicators.

6. The method for detecting hydrogen in transformer oil based on big data analysis and wavelet analysis according to claim 5, characterized in that: The process of hydrogen concentration prediction and mutation point detection is as follows: After the real-time collected data has gone through the same preprocessing and feature extraction process, it is input into the trained SVM model to obtain the predicted hydrogen concentration value; the predicted hydrogen concentration sequence is decomposed using cubic spline wavelet to detect mutation points.

7. Application of the method for detecting hydrogen in transformer oil based on big data analysis and wavelet analysis according to any one of claims 1 to 6 in transformer fault detection.

8. A hydrogen detection system in transformer oil based on big data analysis and wavelet analysis, characterized in that: include: Data acquisition module: collects gas data in transformer oil and performs preprocessing; Data processing module: Use wavelet analysis to smooth the pre-processed data; Feature extraction module: By extracting features from the smoothed data, key parameters related to changes in hydrogen concentration are identified; Model training module: Use the extracted features to train the support vector machine model to establish a relationship model between hydrogen concentration and other variables; And, monitoring module: input the smoothed data into the trained support vector machine model to predict the hydrogen concentration in real time, and combine with wavelet analysis to monitor the mutation points in the hydrogen concentration signal.

9. A computer storage medium storing a readable program, characterized in that: When the program is running, it can execute the method for detecting hydrogen in transformer oil based on big data analysis and wavelet analysis as described in any one of claims 1 to 6.

10. An electronic device, characterized in that: include: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the method for detecting hydrogen in transformer oil based on big data analysis and wavelet analysis as described in any one of claims 1 to 6.

Images