Transformer oil chromatography fault detection method and system based on Raman technology
By using Raman technology and support vector machine model in the transformer oil chromatography fault detection, the problems of insufficient accuracy, poor adaptability and low real-time in the existing technology are solved, and higher detection accuracy and real-time are achieved.
Patent Information
- Application Number
- CN202411536150.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-31
- Publication Date
- 2025-03-14
AI Technical Summary
The prior art has problems such as insufficient accuracy, poor adaptability and low real-time performance in transformer oil chromatography fault detection, which is difficult to meet the high requirements for transformer fault detection in power systems.
The transformer oil chromatography fault detection method based on Raman technology is adopted. By collecting, filtering and purifying the transformer oil, data is collected using a Raman spectrometer, and the data is preprocessed, and the support vector machine model is trained to achieve fault detection.
It improves the accuracy and real-time performance of chromatographic fault detection of transformer oil, and can more effectively identify subtle chemical composition changes in transformer oil, supporting early fault diagnosis.
Smart Images
Figure CN119619331A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of transformer oil chromatographic fault detection, and in particular relates to a transformer oil chromatographic fault detection method and system based on Raman technology. Background Art
[0002] In the power system, transformers are key power equipment, and monitoring their health status and detecting faults are crucial. Transformer oil is the insulation and cooling medium inside the transformer, and the changes in its chromatographic composition are directly related to the operating status of the transformer. Therefore, it is urgent to develop an efficient and accurate transformer oil chromatographic fault detection technology to prevent faults and ensure the stable operation of the power system.
[0003] Experts have conducted special research on transformer health status monitoring and fault detection for a long time. By adopting the enhanced gray wolf algorithm, the energy error is used as the fitness function, and then the energy value of each characteristic modal component is decomposed and calculated to form a characteristic data set to characterize the operating state of the transformer. The differential evolution algorithm is combined with the wavelet neural network to design a neural network structure and sample based on state data for distribution transformer fault detection. By artificially simulating the fault and normal operating states of the transformer and using the k-means SMOTE data expansion method, a limited number of fault data sets can be expanded to make the amount of fault data consistent with the amount of normal data.
[0004] Although the above research has made significant progress in the field of transformer fault detection, with the continuous expansion of the scale of power systems and the increasing complexity of the operating environment, higher requirements have been put forward for transformer fault detection technology, including higher accuracy, stronger adaptability, and better real-time performance. Therefore, a transformer oil chromatography fault detection method based on Raman technology is proposed. Summary of the invention
[0005] In view of the deficiencies in the prior art, the purpose of the present invention is to provide a transformer oil chromatography fault detection method and system based on Raman technology, which solves the problems in the prior art.
[0006] The purpose of the present invention can be achieved by the following technical solutions:
[0007] A transformer oil chromatographic fault detection method based on Raman technology comprises the following steps:
[0008] Collect transformer oil, filter it and purify it;
[0009] Raman spectrometer is used to collect Raman spectrum data of transformer oil;
[0010] Preprocessing the collected Raman spectrum data;
[0011] The support vector machine model is used as the transformer oil chromatographic fault detection model, and the preprocessed Raman spectroscopy data is used to train the support vector machine model.
[0012] Transformer oil fault detection is performed using the trained support vector machine model.
[0013] Furthermore, the filtering treatment is: filtering suspended particles and impurities in the transformer oil with a filter; and the purification treatment is: purifying the transformer oil with an inert gas to remove dissolved gas in the oil sample.
[0014] Furthermore, the Raman spectrometer uses a two-dimensional charge-coupled device array detector with a pixel size of 1044 x 64, a signal-to-noise ratio of up to 1000:1, and a dynamic range of 85000:1; the spectral range is 785nm to 1010nm, and the Raman shift range is 0 to 2850cm -1 .
[0015] Furthermore, the resolution of the Raman spectrometer is less than 4 cm -1 ; The laser wavelength in the Raman spectrometer is 785nm, the output power is 500MW, and the typical value is 375MW.
[0016] Furthermore, the data acquisition parameters in the Raman spectrometer include an adjustable integration time range from 8 milliseconds to 60 minutes; and the analog-to-digital conversion resolution of the Raman spectrometer is 18 bits.
[0017] Furthermore, the preprocessing of the Raman spectroscopy data includes: background subtraction and noise filtering;
[0018] Among them, the background subtraction formula for Raman spectrum data is:
[0019] I sample =I measured -I background
[0020] In the formula, I sample is the sample spectrum intensity after background subtraction, I measured To test the spectral intensity, I background is the blank spectrum background intensity;
[0021] The formula for noise filtering of Raman spectroscopy data is:
[0022] I denoised (t) = H low (t)+H high (t)·n(t)
[0023] In the formula, I denoised (t) is the spectral intensity denoised by wavelet transform, Hlow (t) is the low-frequency component at time t, representing the main trend of the signal, H high (t) is the high frequency component, and n(t) is the noise of the original spectrum.
[0024] Furthermore, the steps of training the support vector machine model are:
[0025] S41, dividing the preprocessed Raman spectrum data into a training set and a test set;
[0026] S42, using radial basis function as kernel function, through grid search and 5-fold cross validation, using the training set to train and optimize the key parameters of the support vector machine model, including penalty parameter C and kernel function parameters;
[0027] S43, use the test set data to evaluate the generalization ability of the support vector machine model.
[0028] A transformer oil chromatography fault detection system based on Raman technology, comprising:
[0029] Sample collection and processing module: collects transformer oil and performs filtering and purification;
[0030] Spectral data acquisition module: uses Raman spectrometer to collect Raman spectrum data of transformer oil;
[0031] Spectral data processing module: pre-process the collected Raman spectral data;
[0032] Model training module: The support vector machine model is used as the transformer oil chromatographic fault detection model, and the pre-processed Raman spectrum data is used to train the support vector machine model;
[0033] And, fault detection module: use the trained support vector machine model to detect transformer oil faults.
[0034] A computer storage medium stores a readable program, which can execute the above-mentioned transformer oil chromatographic fault detection method based on Raman technology when the program is running.
[0035] 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;
[0036] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned transformer oil chromatography fault detection method based on Raman technology.
[0037] Beneficial effects of the present invention:
[0038] 1. The present invention uses Raman spectrometer to conduct detailed analysis on transformer oil samples, optimizes sample collection and processing procedures, ensures high-quality data input, and improves the accuracy of fault detection.
[0039] 2. The present invention improves the quality of spectral data through background subtraction and noise filtering technology, providing reliable data support for the model.
[0040] 3. The SVM model adopted in the present invention has the advantages of accuracy, high efficiency, low latency, etc., and is more suitable for transformer oil chromatography fault detection systems with high real-time requirements. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] 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.
[0042] Figure 1 This is a comparison chart of the detection accuracy of the three algorithms of GBM, RF and SVM of the present invention;
[0043] Figure 2 This is a comparison chart of the detection time of the three algorithms of GBM, RF and SVM of the present invention;
[0044] Figure 3 It is a comparison diagram of the detection time delay of the three algorithms of GBM, RF and SVM of the present invention. DETAILED DESCRIPTION
[0045] 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.
[0046] Example 1
[0047] A transformer oil chromatographic fault detection method based on Raman technology comprises the following steps:
[0048] S1, collecting transformer oil, filtering and purifying it;
[0049] The oil samples of the transformer are collected using an automatic sampler. The oil samples are safely extracted from the oil sampling port of the transformer without introducing external contaminants. During the collection process, the sampling environment is kept clean to prevent the oil samples from being contaminated by air and moisture. The collected oil samples are stored in brown glass bottles, which have been pre-cleaned and dried. The brown glass bottles can effectively block light and reduce the impact of photochemical reactions on the composition of the oil samples. Each sample bottle has a unique label that records the specific sampling time, transformer number, and operating status at the time. After the oil sample is collected, it is necessary to ensure that the sample is kept at a constant temperature and protected from light during the transportation of the sample to the laboratory to maintain the chemical stability of the oil sample.
[0050] After the sample arrives at the laboratory, the suspended particles and impurities in the oil are filtered out using a filter with a suitable pore size. In order to further ensure the accuracy of the analysis, the oil sample is purified with an inert gas, which is continuously introduced into the container containing the oil sample to remove the dissolved gas in the oil sample, reduce the interference of the gas on the Raman spectroscopy analysis, and improve the quality of the spectral signal.
[0051] Finally, before Raman spectroscopy analysis, the sample is transferred to a sample cell specially designed for the spectrometer and the sample surface is ensured to be flat to obtain a uniform and high-quality spectral signal.
[0052] S2, using Raman spectrometer to collect Raman spectrum data of transformer oil;
[0053] The Raman spectrometer uses a two-dimensional charge-coupled device array detector with a pixel size of 1044 x 64, a signal-to-noise ratio of up to 1000:1, and a dynamic range of 85000:1; the spectral range is 785nm to 1010nm, and the Raman shift range is 0 to 2850cm -1 , sufficient to capture the Raman scattering properties of key gases dissolved in oil.
[0054] Raman spectrometer also has a smaller than 4cm -1 The high resolution enables the Raman spectrometer to distinguish very close Raman peaks, thereby accurately identifying and quantifying the dissolved gas components in transformer oil. The laser wavelength in the Raman spectrometer is 785nm, the output power is 500MW, the typical value is 375MW, and the power stability is controlled within ±1%, ensuring the stability and reliability of the excitation source.
[0055] In addition, the data acquisition parameters in the Raman spectrometer include an adjustable integration time range from 8 milliseconds to 60 minutes to adapt to different sample characteristics and analysis requirements; the Raman spectrometer has an analog-to-digital conversion resolution of 18 bits, providing extremely high accuracy for spectral data, so that the spectral data obtained from transformer oil samples can accurately reflect its chemical composition.
[0056] S3, preprocessing the collected Raman spectrum data;
[0057] In this embodiment, the collected Raman spectrum data information is shown in Table 1:
[0058] Table 1 Raman spectrum data
[0059]
[0060] The preprocessing of Raman spectroscopy data includes: background subtraction, noise filtering, removal of duplicate data, difference processing, etc. Among them, the background subtraction formula for Raman spectroscopy data is:
[0061] I sample =I measured -I background
[0062] In the formula, I sample is the sample spectrum intensity after background subtraction, I measured To test the spectral intensity, I background is the blank spectrum background intensity.
[0063] The formula for noise filtering of Raman spectroscopy data is:
[0064] I denoised (t) = H low (t)+H high (t)·n(t)
[0065] In the formula, I denoised (t) is the spectral intensity denoised by wavelet transform, H low (t) is the low-frequency component at time t, representing the main trend of the signal, H high (t) is the high frequency component, and n(t) is the noise of the original spectrum.
[0066] Remove duplicate data from Raman spectroscopy data: Create a pandasDataFrame containing Raman data, and use Python software programming to sort column data to remove duplicate data points.
[0067] The Raman spectral data is subjected to difference processing: the baseline of the spectral data is corrected before the difference processing, and the selected spectrally smooth reference spectrum is subtracted from the original data to obtain the difference spectrum.
[0068] S4, using the support vector machine (SVM) model as the transformer oil chromatographic fault detection model, and using the preprocessed Raman spectroscopy data to train the support vector machine model;
[0069] The transformer oil chromatographic fault detection model focuses on feature extraction. It analyzes the preprocessed Raman spectral data to identify the Raman peak wavenumber, peak intensity and half-maximum full width of specific gas molecules directly related to transformer oil faults. Feature selection is achieved through recursive feature elimination to ensure that the selected features have statistically significant fault classification and discrimination capabilities.
[0070] The steps to train a support vector machine model are:
[0071] S41, dividing the preprocessed Raman spectrum data into a training set and a test set;
[0072] The training set is used to train the support vector machine model, and the test set is used to evaluate the generalization ability of the support vector machine model.
[0073] S42, radial basis function is used as kernel function, and the key parameters of the support vector machine model, including penalty parameter C and kernel function parameters, are trained and optimized using the training set through grid search and 5-fold cross validation.
[0074] S43, use the test set data to evaluate the generalization ability of the support vector machine model.
[0075] Finally, the model was validated on an independent validation set to verify its accuracy and stability in practical applications.
[0076] S5, using the trained support vector machine model to perform fault detection on transformer oil.
[0077] Example 2
[0078] In order to verify the effectiveness of the SVM model in transformer oil spectrum fault detection, the SVM is compared with the random forest (RF) and gradient propulsion machine (GBM) models. The experimental process is as follows: the SVM parameters are optimized through grid search and cross-validation of the penalty parameter C and the kernel function parameters, and RF and GBM are adjusted to operate in the optimal state. Finally, the applicability of the support vector machine in the fault detection task is evaluated through performance comparison and analysis, and the advantages and limitations of the support vector machine compared with other algorithms are discussed.
[0079] Experimental data such as Figures 1 to 3 As shown:
[0080] Figure 1In the figure, the triangle represents the detection accuracy of 30 sets of Raman data under the GBM algorithm, the square represents the detection accuracy of 30 sets of Raman data under the RF algorithm, and the circle represents the detection accuracy of 30 sets of Raman data under the SVM algorithm; the results show that the accuracy of SVM in transformer oil spectrum fault detection is significantly higher than that of RF and GBM, both above 95.14%, and up to 99.99%. This significant performance advantage shows that SVM has extremely high classification accuracy when processing such complex spectral data. In addition, the high accuracy of the support vector machine also reflects its sensitivity in identifying subtle changes in chemical composition in transformer oil, which is crucial for early fault diagnosis.
[0081] Figure 2 The detection time data shows the efficiency of the three algorithms in transformer oil chromatographic fault detection. After detailed analysis, it was found that the detection time of SVM was between 501ms and 646ms, while the detection time of RF and GBM was between 650ms and 749ms and 754ms and 933ms respectively. This difference in efficiency is related to the computational complexity of the algorithm.
[0082] Figure 3 It shows that: SVM has the lowest detection delay, the lowest is 106ms, followed by RF, the lowest is 256ms, and GBM has the highest detection delay, the maximum is 746ms. The evaluation results of SVM are significantly better than RF and GBM, indicating that SVM has extremely high efficiency and fast data processing capabilities in fault detection.
[0083] 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.
[0084] 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 transformer oil chromatographic fault detection method based on Raman technology, characterized in that: The following steps are involved: Collect transformer oil, filter it and purify it; Raman spectrometer is used to collect Raman spectrum data of transformer oil; Preprocessing the collected Raman spectrum data; The support vector machine model is used as the transformer oil chromatographic fault detection model, and the preprocessed Raman spectroscopy data is used to train the support vector machine model. Transformer oil fault detection is performed using the trained support vector machine model.
2. The transformer oil chromatographic fault detection method based on Raman technology according to claim 1 is characterized in that: The filtering process is to filter the suspended particles and impurities in the transformer oil with a filter; the purification process is to purify the transformer oil with an inert gas to remove the dissolved gas in the oil sample.
3. The transformer oil chromatographic fault detection method based on Raman technology according to claim 1 is characterized in that: The Raman spectrometer uses a two-dimensional charge-coupled device array detector with a pixel size of 1044x64, a signal-to-noise ratio of up to 1000:1, and a dynamic range of 85000:1; the spectral range is 785nm to 1010nm, and the Raman shift range is 0 to 2850cm -1 .
4. The transformer oil chromatographic fault detection method based on Raman technology according to claim 3 is characterized in that: The resolution of the Raman spectrometer is less than 4 cm -1 ; The laser wavelength in the Raman spectrometer is 785nm, the output power is 500MW, and the typical value is 375MW.
5. The transformer oil chromatographic fault detection method based on Raman technology according to claim 4 is characterized in that: The data acquisition parameters in the Raman spectrometer include an adjustable integration time range from 8 milliseconds to 60 minutes; the analog-to-digital conversion resolution of the Raman spectrometer is 18 bits.
6. The transformer oil chromatography fault detection method based on Raman technology according to claim 1 is characterized in that: The preprocessing of the Raman spectrum data includes: background subtraction and noise filtering; Among them, the background subtraction formula for Raman spectrum data is: I sample =I measured -I background In the formula, I sample is the sample spectrum intensity after background subtraction, I measured To test the spectral intensity, I background is the blank spectrum background intensity; The formula for noise filtering of Raman spectroscopy data is: I denoised (t)=H low (t)+H high (t)·n(t) In the formula, I denoised (t) is the spectral intensity denoised by wavelet transform, H low (t) is the low-frequency component at time t, representing the main trend of the signal, H high (t) is the high frequency component, and n(t) is the noise of the original spectrum.
7. The transformer oil chromatographic fault detection method based on Raman technology according to claim 1 is characterized in that: The steps of training the support vector machine model are: S41, dividing the preprocessed Raman spectrum data into a training set and a test set; S42, using radial basis function as kernel function, through grid search and 5-fold cross validation, using the training set to train and optimize the key parameters of the support vector machine model, including penalty parameter C and kernel function parameters; S43, use the test set data to evaluate the generalization ability of the support vector machine model.
8. A transformer oil chromatography fault detection system based on Raman technology, characterized in that: include: Sample collection and processing module: collects transformer oil and performs filtering and purification; Spectral data acquisition module: uses Raman spectrometer to collect Raman spectrum data of transformer oil; Spectral data processing module: pre-process the collected Raman spectral data; Model training module: The support vector machine model is used as the transformer oil chromatographic fault detection model, and the pre-processed Raman spectrum data is used to train the support vector machine model; And, fault detection module: use the trained support vector machine model to detect transformer oil faults.
9. A computer storage medium storing a readable program, characterized in that: When the program is running, it can execute the transformer oil chromatography fault detection method based on Raman technology as described in any one of claims 1 to 7.
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 transformer oil chromatography fault detection method based on Raman technology as described in any one of claims 1-7.