CN116010884A - Fault diagnosis method of SSA-LightGBM oil-immersed transformer based on principal component analysis - Google Patents
Fault diagnosis method of SSA-LightGBM oil-immersed transformer based on principal component analysis Download PDFInfo
- Publication number
- CN116010884A CN116010884A CN202211461338.8A CN202211461338A CN116010884A CN 116010884 A CN116010884 A CN 116010884A CN 202211461338 A CN202211461338 A CN 202211461338A CN 116010884 A CN116010884 A CN 116010884A
- Authority
- CN
- China
- Prior art keywords
- transformer
- characteristic
- fault diagnosis
- lightgbm
- matrix
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Testing Electric Properties And Detecting Electric Faults (AREA)
Abstract
The invention discloses a fault diagnosis method of an SSA-LightGBM oil-immersed transformer based on principal component analysis, which comprises the following steps: collecting historical sample data of transformer fault characteristic gas, determining the type of the transformer fault, and adopting a non-coding ratio method to obtain the real-time running characteristic of the transformer as a parameter to be measured to construct a matrix to be measured; extracting characteristic parameters of the matrix to be detected by using a principal component analysis method, and constructing a characteristic sample matrix by using the characteristic parameters; and constructing a LightGBM model, determining super parameters and parameter selection ranges which need to be optimized, optimizing the super parameters of the LightGBM model by utilizing a sparrow search algorithm, constructing an SSA-LightGBM fault diagnosis model, and inputting the characteristic sample matrix into the trained fault diagnosis model for analysis to obtain a fault diagnosis result. The transformer fault diagnosis method has the advantages of high accuracy of the transformer fault diagnosis result, high model diagnosis speed and smaller memory occupied by the model.
Description
Technical Field
The invention relates to the field of fault diagnosis and research of transformers, in particular to a fault diagnosis method of an SSA-LightGBM oil-immersed transformer based on principal component analysis.
Background
The transformer is core equipment of the power system, and potential faults are found in time to play an important role in safety and stability of the power system. The oil immersed transformer can generate gases such as hydrogen, methane, ethylene, acetylene, carbon monoxide, carbon dioxide and the like due to the cracking of insulating oil and solid insulation caused by long-time operation, when the transformer has different types of faults, specific gas components can be increased, the fault types of the transformer and the gas component changes are closely related, and a common method is an analysis method of dissolved gas in oil (DissolvedCasesAnalysis, DGA). The fault diagnosis method based on DGA generally comprises algorithms such as an Artificial Neural Network (ANN), a Support Vector Machine (SVM), a Fuzzy Theory (FT) and the like, so that the accuracy of fault diagnosis is greatly improved. However, these methods have some drawbacks, in which: the fuzzy theory has the problem of insufficient learning ability; the artificial neural network model improves the fault diagnosis precision of the transformer, but the artificial neural network model needs a large amount of forgetting data to train, and in actual engineering, the fault data which can be used for training is very limited and is easy to sink into local optimum; the SVM can be used for fault diagnosis under the condition of small sample data size, and compared with an artificial neural network, the SVM can better process local minima and has stronger learning generalization capability, but incorrect parameter values can cause larger errors on diagnosis results.
In view of the above, an SSA-LightGBM oil immersed transformer fault diagnosis method based on principal component analysis is provided, which can accelerate training speed, occupy small memory, and stably improve accuracy of transformer fault diagnosis.
Disclosure of Invention
The invention aims to provide a fault diagnosis method of an SSA-LightGBM oil immersed transformer based on principal component analysis, which realizes accurate judgment of the fault type of the transformer.
The invention is realized by adopting the following technical scheme: a fault diagnosis method for an SSA-LightGBM oil-immersed transformer based on principal component analysis comprises the following steps:
step 1: collecting historical sample data of transformer fault characteristic gas, determining the type of the transformer fault, and adopting a non-coding ratio method to obtain real-time running of the transformer as a parameter to be tested to construct a matrix to be tested;
step 2: carrying out normalization pretreatment on the parameters to be detected, extracting characteristic parameters of the matrix to be detected by using a principal component analysis method, and constructing a characteristic sample matrix by using the characteristic parameters;
step 3: and constructing a LightGBM model, determining super parameters and parameter selection ranges which need to be optimized, optimizing the super parameters of the LightGBM model by utilizing a sparrow search algorithm, constructing an SSA-LightGBM fault diagnosis model, and inputting the characteristic sample matrix into the trained SSA-LightGBM fault diagnosis model for analysis to obtain a fault diagnosis result.
Preferably, transformer fault signature gas history sample data is collected and the transformer fault type is determined:
step 1: the transformer fault signature gas comprises:
H 2 、CH 4 、C 2 H 2 、C 2 H 4 、C 2 H 6 、CO、CO 2 ；
step 2: the transformer fault type includes: high-temperature thermal faults, low-temperature thermal faults, medium-temperature thermal faults, high-energy discharge, low-energy discharge, partial discharge and normal states.
Preferably, the method for obtaining the real-time running of the transformer by using the non-coding ratio method as the parameter to be measured to construct the matrix to be measured includes:
Preferably, the normalizing pretreatment for the parameter to be measured includes:
step 1: calculating the mean value and standard deviation of the parameters to be measured;
step 2: calculating to obtain a normalized value of each parameter to be measured according to the mean value and standard deviation of the parameter to be measured;
step 3: and constructing the matrix to be tested according to the normalized value of each parameter to be tested.
Preferably, the extracting the feature parameters of the matrix to be tested by using a principal component analysis method includes:
step 1: calculating a covariance matrix of the parameters to be measured in the matrix to be measured;
step 2: calculating eigenvalues of the covariance matrix and corresponding orthogonalization unit eigenvectors;
step 3: selecting a principal component according to the eigenvalue and the orthogonalization unit eigenvector;
step 4: and calculating the load of the main component, and obtaining the characteristic parameters according to the load of the main component to construct a characteristic matrix.
Preferably, constructing the LightGBM model comprises:
step 1: constructing a data set and category characteristics in the parameters to be measured, and calculating an initial gradient value;
step 2: constructing a decision tree according to the initial gradient value, and establishing a histogram;
step 3: calculating splitting income according to the histogram, and selecting the optimal splitting characteristic to obtain a splitting threshold;
step 4: and (3) establishing root nodes, repeating the steps until the limit of the number of leaves is reached or all the leaf nodes can not be continuously segmented, and updating the gradient values of the tree to complete the construction of all the trees.
Preferably, determining the super parameter and the parameter selection range to be optimized includes:
step 1: the super-parameter learning rate is set to be in a range of [0.1,0.5,0.8];
step 2: setting the value range of the super parameter max_depth as a section [16,32,64];
step 3: the super parameter subsamples are set to have a value range of interval [0.1,1];
step 4: the super parameter colsample_byte is set to have a value range of interval [0, 1].
Preferably, optimizing the LightGBM model super parameters using a sparrow search algorithm includes:
step 1: extracting characteristic quantity of fault data as an input signal of a model, and outputting the characteristic quantity as a fault type of the transformer;
step 2: initializing a sparrow search algorithm and a LightGBM related super parameter, and setting a population scale and a maximum iteration number;
step 3: taking the accuracy of cross verification as the fitness value of the sparrow individual;
step 4: generating a reverse sparrow population, sorting the fitness of all the sparrow population individuals, wherein the individuals with higher fitness values are discoverers, and the rest are joiners;
step 5: updating the position of the finder, wherein sparrow can be widely searched when the finder is in a safe state, and if the sparrow is larger than the early warning value, the population has anti-predation behavior;
step 6: updating the position of the subscriber according to the sorting principle, when
When the individual fitness value is relatively low, the user needs to search other positions to improve the individual fitness;
step 7: and judging whether the fitness value meets the termination condition, and if so, exiting the cyclic sparrow optimizing algorithm to finish to obtain the optimal parameters. And (3) constructing an SSA-LightGBM fault diagnosis model, otherwise, returning to the steps, and finally meeting the termination condition to obtain the optimal parameters.
Preferably, the training of the SSA-LightGBM fault diagnosis model includes:
step 1: collecting historical sample data of transformer fault characteristic gas, determining the type of the transformer fault, and adopting a non-coding ratio method to obtain real-time running of the transformer as a parameter to be tested to construct a matrix to be tested;
step 2: carrying out normalization pretreatment on the parameters to be detected, extracting characteristic parameters of the matrix to be detected by using a principal component analysis method, and constructing a characteristic sample matrix by using the characteristic parameters;
step 3: and constructing a LightGBM model, determining super parameters and parameter selection ranges which need to be optimized, optimizing the super parameters of the LightGBM model by utilizing a sparrow search algorithm, constructing a fault diagnosis model, and training the fault diagnosis model by utilizing the initial characteristic sample matrix to obtain an SSA-LightGBM fault diagnosis model.
Preferably, inputting the feature sample matrix into a trained SSA-LightGBM fault diagnosis model for analysis, to obtain a fault diagnosis result, including:
step 1: collecting the latest transformer fault characteristic gas, and obtaining the real-time running characteristic construction test sample matrix of the transformer by adopting a non-coding ratio method;
step 2: carrying out normalization pretreatment on the test sample matrix, and multiplying the test sample matrix by the characteristic sample matrix obtained after the dimension reduction by using a principal component analysis method to obtain a test characteristic matrix;
step 3: inputting the test feature matrix into a trained SSA-LightGBM fault diagnosis model for analysis, and obtaining a fault diagnosis result.
The invention relates to a fault diagnosis method of an SSA-LightGBM oil immersed transformer based on principal component analysis, which comprises the following steps: collecting historical sample data of transformer fault characteristic gas, determining the type of the transformer fault, and adopting a non-coding ratio method to obtain the real-time running characteristic of the transformer as a parameter to be measured to construct a matrix to be measured; extracting characteristic parameters of the matrix to be detected by using a principal component analysis method, and constructing a characteristic sample matrix by using the characteristic parameters; and constructing a LightGBM model, determining super parameters and parameter selection ranges which need to be optimized, optimizing the super parameters of the LightGBM model by utilizing a sparrow search algorithm, constructing an SSA-LightGBM fault diagnosis model, and inputting the characteristic sample matrix into the trained SSA-LightGBM fault diagnosis model for analysis to obtain a fault diagnosis result.
Therefore, the transformer fault diagnosis model provided by the application is based on the principal component analysis method and the LightGBM model, the principal component analysis method is utilized to conduct feature extraction on the dissolved gas parameters in the transformer oil, the feature dimension is reduced, the SSA-LightGBM fault diagnosis model is built by optimizing the super parameters of the LightGBM model through the sparrow optimizing algorithm, and the training sample is input into the model to obtain the transformer fault result.
Drawings
Figure 1 is a block flow diagram of a method according to the present invention.
Fig. 2 is a block flow diagram of a sparrow optimization algorithm of an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more clear, the present invention will be described in further detail with reference to the accompanying drawings and specific embodiments. The specific embodiments described herein are to be considered in an illustrative sense only and are not intended to limit the invention.
The embodiment provides a fault diagnosis method for an SSA-LightGBM oil-immersed transformer based on principal component analysis, as shown in fig. 1, which is a flowchart of the fault diagnosis method for the SSA-LightGBM oil-immersed transformer based on principal component analysis.
The fault diagnosis method of the SSA-LightGBM oil-immersed transformer based on principal component analysis, provided by the embodiment, comprises the following steps:
s101: collecting transformer fault characteristic gas history sample data and determining the type of transformer faults:
specifically, a DGA on-line monitoring system is adopted to collect the gas history sample data of the fault characteristics of the transformer,
the target dissolved gas includes: h 2 、CH 4 、C 2 H 2 、C 2 H 4 、C 2 H 6 、CO、CO 2 。
The above-mentioned transformer fault types include: high-temperature thermal faults, low-temperature thermal faults, medium-temperature thermal faults, high-energy discharge, low-energy discharge, partial discharge and normal states.
S102: the method comprises the steps of obtaining the gas concentration of the transformer running in real time by adopting a coding-free ratio method as a parameter to be measured, and constructing a matrix to be measured:
on the basis of extracting the historical sample data of the transformer fault characteristic gas, a non-coding ratio method is adopted to obtain the real-time running of the transformer as a parameter to be tested to construct a matrix to be tested.
the above example of the transformer fault signature gas history sample data is only one implementation provided in the present application, and is not limited only, and may be differently set according to actual needs for the type of the target dissolved gas.
S103: and carrying out normalization pretreatment on the parameters to be measured:
in one specific embodiment, calculating the mean value and standard deviation of the parameters to be measured; calculating to obtain a normalized value of each parameter to be measured according to the mean value and standard deviation of the parameter to be measured; and constructing the matrix to be tested according to the normalized value of each parameter to be tested.
Specifically, after the concentration of each target dissolved gas in the transformer oil is obtained as a parameter to be measured, the normalization treatment is performed by adopting a Z-score method, and the formula is expressed as follows:
wherein x is * The characteristic quantity obtained after normalization, x is the original data,
the mean value of the original data, sigma is the standard deviation of the original data.
S104: extracting characteristic parameters of the matrix to be detected by using a principal component analysis method, and constructing a characteristic sample matrix by using the characteristic parameters:
in a specific embodiment, the extracting the feature parameters of the matrix to be tested by using a principal component analysis method includes: calculating a covariance matrix of the parameters to be measured in the matrix to be measured; calculating eigenvalues of the covariance matrix and corresponding orthogonalization unit eigenvectors; selecting a principal component according to the eigenvalue and the orthogonalization unit eigenvector; and calculating the load of the main component, and obtaining the characteristic parameters according to the load of the main component.
Specifically, a covariance matrix of parameters to be measured is calculated:
ε＝(s ij ) m×m ；
wherein, the liquid crystal display device comprises a liquid crystal display device,
further, the eigenvalue lambda of the covariance matrix is calculated i Corresponding orthogonalization unit feature vector a i . The first p larger eigenvalues λ of ε 1 ≥λ 2 ...λ p > 0, the variance corresponding to the first p principal components, lambda i Corresponding unit feature vector a i Is the principal component analysis F i With respect to the coefficient of the original variable, the i-th principal component F of the original variable i The method comprises the following steps: f (F) i ＝a i Y, the variance contribution ratio of the principal component is used to reflect the information amount size,
finally, several main components, namely F, are selected 1 ,F 2 ,...,F p The determination of P in (c) is determined by variance accumulation contribution G (P):
when the contribution rate is more than 95%, the selected principal component can reflect the information of the original variable, the corresponding p is the first p principal components extracted, F 1 Marked as the first main component, F 2 Is marked as a second main component, …, F p The p-th principal component.
Further, the principal component load is a function of principal component F i With the original variable Y j Degree of correlation between the original variable Y j (j=1, 2,., m) in main component F i (i=1, 2,., p) ij (i=1, 2., p; j=1, 2., m) is:
calculating the scores of the initial data to be tested on p principal components:
F i ＝l 1i Y 1 +l 2i Y 2 +....+l mi Y m (i＝1,2,...,p)
F 1 ,F 2 ,...,F p namely, the original variable index:
Y 1 ,Y 2 ,...,Y m
the new variable obtained after principal component analysis is used for constructing a characteristic sample matrix as follows:
s105: building a LightGBM model:
specifically, the process of constructing the tree by the LightGBM is as follows: let the dataset be s= { (x) i ,y i )i＝1,2,...,n},x i ＝{X i1 ,X i2 ,...,X in -where m is the number of features, y i Is a category feature.
Further, an initial gradient value is calculated for the characteristic parameter in S:
further, a histogram is constructed:
further, calculating split benefits based on the histogram, and selecting an optimal split characteristic G to obtain a split threshold I:
further, a root node is established:
t＝arg max(G i ),1≤i≤m
node＝(t,G t ,I t )
further, repeating the steps until the limit of the number of leaves is reached or all the leaf nodes can not be continuously segmented, and finally updating the gradient value of the tree to complete the construction of all the trees.
S106: super parameters to be optimized and parameter selection ranges:
specifically, the value range of the super parameter setting is as follows:
the super-parameter learning rate is set to be in a range of [0.1,0.5,0.8];
setting the value range of the super parameter max_depth as a section [16,32,64];
the super parameter subsamples are set to have a value range of interval [0.1,1];
the super parameter colsample_byte is set to have a value range of interval [0, 1].
S107: sparrow search algorithm optimizes the super parameters of the LightGBM model:
specifically, extracting characteristic quantity of fault data as an input signal of a model, and outputting the characteristic quantity as a fault type of a transformer; initializing a sparrow search algorithm and a LightGBM related super parameter, and setting a population scale and a maximum iteration number; taking the accuracy of cross verification as the fitness value of the sparrow individual; generating a reverse sparrow population, sorting the fitness of all the sparrow population individuals, wherein the individuals with higher fitness values are discoverers, and the rest are joiners; updating the position of the finder, wherein sparrow can be widely searched when the finder is in a safe state, and if the sparrow is larger than the early warning value, the population has anti-predation behavior; updating the position of the joiner according to the ordering principle at the time
The individual fitness value is low, and the subscriber needs to search other positions to improve the individual fitness; and judging whether the fitness value meets the termination condition, and if so, exiting the cyclic sparrow optimizing algorithm to finish to obtain the optimal parameters. Construction of SSA-LightGBM-based fault diagnosis modelOtherwise, returning to the steps, and finally meeting the termination condition to obtain the optimal parameters.
Wherein the SSA algorithm parameter settings are shown in table 1.
Table 1: parameters set by SSA algorithm
S106: training to obtain the SSA-LightGBM fault diagnosis model by the following steps:
specifically, extracting the target gas dissolved concentration in the fault transformer oil, and adopting a non-coding ratio method to obtain real-time running of the transformer as a parameter to be tested to construct a matrix to be tested; carrying out normalization pretreatment on the parameters to be detected, extracting characteristic parameters of the matrix to be detected by using a principal component analysis method, and constructing a characteristic sample matrix by using the characteristic parameters; constructing a LightGBM model and determining super parameters and parameter selection ranges which need to be optimized; and optimizing the super parameters of the LightGBM model by utilizing a sparrow search algorithm, constructing a fault diagnosis model, and training the fault diagnosis model by utilizing the initial characteristic sample matrix to obtain the SSA-LightGBM fault diagnosis model.
S106: inputting the characteristic sample matrix into a trained SSA-LightGBM fault diagnosis model for analysis, and obtaining a fault diagnosis result:
collecting the latest transformer fault characteristic gas, and obtaining the real-time running characteristic construction test sample matrix of the transformer by adopting a non-coding ratio method; carrying out normalization pretreatment on the test sample matrix, and multiplying the test sample matrix by the characteristic sample matrix obtained after the dimension reduction by using a principal component analysis method to obtain a test characteristic matrix; inputting the test feature matrix into a trained SSA-LightGBM fault diagnosis model for analysis, and obtaining a fault diagnosis result.
In this embodiment, feature gas is collected during 513 transformer faults, test data is directly input into a trained SSA-LightGBM fault diagnosis model, and classification results are shown in the following table:
TABLE 2 sample distribution
TABLE 3 diagnostic results
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art will understand that; the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced with equivalents; such modifications and substitutions do not depart from the spirit of the invention, which is defined by the following claims.
Claims (10)
1. The fault diagnosis model method for the SSA-LightGBM oil immersed transformer based on principal component analysis is characterized by comprising the following steps of:
step 1: collecting historical sample data of transformer fault characteristic gas, determining the type of the transformer fault, and adopting a non-coding ratio method to obtain real-time running of the transformer as a parameter to be tested to construct a matrix to be tested;
step 2: carrying out normalization pretreatment on the parameters to be detected, extracting characteristic parameters of the matrix to be detected by using a principal component analysis method, and constructing a characteristic sample matrix by using the characteristic parameters;
step 3: and constructing a LightGBM model, determining super parameters and parameter selection ranges which need to be optimized, optimizing the super parameters of the LightGBM model by utilizing a sparrow search algorithm, constructing an SSA-LightGBM fault diagnosis model, and inputting the characteristic sample matrix into the trained SSA-LightGBM fault diagnosis model for analysis to obtain a fault diagnosis result.
2. The transformer fault diagnosis method of step 1 according to claim 1, wherein: collecting transformer fault characteristic gas history sample data and determining the type of transformer faults:
step 1: the transformer fault signature gas comprises: h 2 、CH 4 、C 2 H 2 、C 2 H 4 、C 2 H 6 、CO、CO 2 。
Step 2: the transformer fault type includes: high-temperature thermal faults, low-temperature thermal faults, medium-temperature thermal faults, high-energy discharge, low-energy discharge, partial discharge and normal states.
3. The transformer fault diagnosis method of step 1 according to claim 1, wherein: the method for obtaining the real-time running of the transformer by adopting the non-coding ratio method as the parameter to be tested to construct the matrix to be tested comprises the following steps:
wherein, the method is obtained by using a coding-free ratio method:
4. The transformer fault diagnosis method of step 2 according to claim 1, wherein: carrying out normalization pretreatment on the parameters to be measured:
step 1: calculating the mean value and standard deviation of the parameters to be measured;
step 2: calculating to obtain a normalized value of each parameter to be measured according to the mean value and standard deviation of the parameter to be measured;
step 3: and constructing the matrix to be tested according to the normalized value of each parameter to be tested.
5. The method for diagnosing a transformer fault as recited in claim 1, wherein the extracting the characteristic parameters of the matrix to be tested by using a principal component analysis method comprises:
step 1: calculating a covariance matrix of the parameters to be measured in the matrix to be measured;
step 2: calculating eigenvalues of the covariance matrix and corresponding orthogonalization unit eigenvectors;
step 3: selecting a principal component according to the eigenvalue and the orthogonalization unit eigenvector;
step 4: and calculating the load of the main component, and obtaining the characteristic parameters to construct a characteristic matrix according to the load of the main component.
6. The method for diagnosing a transformer fault in step 3 as claimed in claim 1, wherein constructing the LightGBM model comprises:
step 1: constructing a data set and category characteristics in the parameters to be measured, and calculating an initial gradient value;
step 2: constructing a decision tree according to the initial gradient value, and establishing a histogram;
step 3: calculating splitting income according to the histogram, and selecting the optimal splitting characteristic to obtain a splitting threshold;
step 4: and (3) establishing root nodes, repeating the steps until the limit of the number of leaves is reached or all the leaf nodes can not be continuously segmented, and updating the gradient values of the tree to complete the construction of all the trees.
7. The method for diagnosing a transformer fault as recited in claim 1, wherein determining the hyper-parameters and the selected ranges of parameters to be optimized comprises:
step 1: the super-parameter learning rate is set to be in a range of [0.1,0.5,0.8];
step 2: setting the value range of the super parameter max_depth as a section [16,32,64];
step 3: the super parameter subsamples are set to have a value range of interval [0.1,1];
step 4: the super parameter colsample_byte is set to have a value range of interval [0, 1].
8. The method of step 3 transformer fault diagnosis according to claim 1, wherein optimizing the LightGBM model super-parameters using a sparrow search algorithm comprises:
step 1: extracting characteristic quantity of fault data as an input signal of a model, and outputting the characteristic quantity as a fault type of the transformer;
step 2: initializing a sparrow search algorithm and a LightGBM related super parameter, and setting a population scale and a maximum iteration number;
step 3: taking the accuracy of cross verification as the fitness value of the sparrow individual;
step 4: generating a reverse sparrow population, sorting the fitness of all the sparrow population individuals, wherein the individuals with higher fitness values are discoverers, and the rest are joiners;
step 5: updating the position of the finder, wherein sparrow can be widely searched when the finder is in a safe state, and if the sparrow is larger than the early warning value, the population has anti-predation behavior;
step 6: updating the position of the subscriber according to the sorting principle, when
When the individual fitness value is relatively low, the user needs to search other positions to improve the individual fitness;
step 7: and judging whether the fitness value meets the termination condition, if so, exiting the cyclic sparrow optimizing algorithm to finish to obtain the optimal parameter, constructing an SSA-LightGBM fault diagnosis model, otherwise, returning to the step, and finally, meeting the termination condition to obtain the optimal parameter.
9. The method of step 3 transformer fault diagnosis according to claim 1, wherein training the SSA-LightGBM fault diagnosis model comprises:
step 1: collecting historical sample data of transformer fault characteristic gas, determining the type of the transformer fault, and adopting a non-coding ratio method to obtain the real-time running characteristic of the transformer as a parameter to be tested to construct a training sample matrix;
step 2: carrying out normalization pretreatment on the training sample matrix, extracting characteristic parameters of the matrix to be tested by using a principal component analysis method, and constructing a characteristic sample matrix by using the characteristic parameters;
step 3: constructing a LightGBM model and determining super parameters and parameter selection ranges which need to be optimized;
step 4: and optimizing the super parameters of the LightGBM model by utilizing a sparrow search algorithm, constructing a fault diagnosis model, and training the fault diagnosis model by utilizing the initial characteristic sample matrix to obtain the SSA-LightGBM fault diagnosis model.
10. The method for diagnosing a fault of a step 3 transformer according to claim 1, wherein inputting the feature sample matrix into a trained SSA-LightGBM fault diagnosis model for analysis, to obtain a fault diagnosis result, comprises:
step 1: collecting the latest transformer fault characteristic gas, and obtaining the real-time running characteristic construction test sample matrix of the transformer by adopting a non-coding ratio method;
step 2: carrying out normalization pretreatment on the test sample matrix, and multiplying the test sample matrix by the characteristic sample matrix obtained after the dimension reduction by using a principal component analysis method to obtain a test characteristic matrix;
step 3: inputting the test feature matrix into a trained SSA-LightGBM fault diagnosis model for analysis, and obtaining a fault diagnosis result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211461338.8A CN116010884A (en) | 2022-11-21 | 2022-11-21 | Fault diagnosis method of SSA-LightGBM oil-immersed transformer based on principal component analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211461338.8A CN116010884A (en) | 2022-11-21 | 2022-11-21 | Fault diagnosis method of SSA-LightGBM oil-immersed transformer based on principal component analysis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116010884A true CN116010884A (en) | 2023-04-25 |
Family
ID=86030702
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211461338.8A Pending CN116010884A (en) | 2022-11-21 | 2022-11-21 | Fault diagnosis method of SSA-LightGBM oil-immersed transformer based on principal component analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116010884A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116561528A (en) * | 2023-05-08 | 2023-08-08 | 重庆市荣冠科技有限公司 | RUL prediction method of rotary machine |
| CN117972616A (en) * | 2024-03-28 | 2024-05-03 | 江西江投能源技术研究有限公司 | Pumped storage generator set safety state monitoring and diagnosing method and system |
-
2022
- 2022-11-21 CN CN202211461338.8A patent/CN116010884A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116561528A (en) * | 2023-05-08 | 2023-08-08 | 重庆市荣冠科技有限公司 | RUL prediction method of rotary machine |
| CN116561528B (en) * | 2023-05-08 | 2024-03-01 | 重庆市荣冠科技有限公司 | RUL prediction method of rotary machine |
| CN117972616A (en) * | 2024-03-28 | 2024-05-03 | 江西江投能源技术研究有限公司 | Pumped storage generator set safety state monitoring and diagnosing method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116010884A (en) | Fault diagnosis method of SSA-LightGBM oil-immersed transformer based on principal component analysis | |
| CN111722046A (en) | Transformer fault diagnosis method based on deep forest model | |
| CN113344288B (en) | Cascade hydropower station group water level prediction method and device and computer readable storage medium | |
| US20240133391A1 (en) | Prediction method for stall and surging of axial-flow compressor based on deep autoregressive network | |
| CN111680726A (en) | Transformer fault diagnosis method and system based on neighbor component analysis and k neighbor learning fusion | |
| CN113889198A (en) | Transformer fault diagnosis method and equipment based on oil chromatogram time-frequency domain information and residual error attention network | |
| CN115577639A (en) | Transformer fault diagnosis method based on sparrow search method optimization support vector machine | |
| CN116595445A (en) | Transformer fault diagnosis method based on random forest feature optimization and improved support vector machine | |
| CN116224112A (en) | XGBoost model-based battery impedance spectrum frequency importance analysis method | |
| CN114358116A (en) | Oil-immersed transformer fault diagnosis method and system and readable storage medium | |
| CN114330486A (en) | Power system bad data identification method based on improved Wasserstein GAN | |
| CN116911634A (en) | Method and device for analyzing lower limit influence factor of class D voltage based on random forest | |
| CN116736214A (en) | CVT stability evaluation method and system | |
| CN116561569A (en) | Industrial power load identification method based on EO feature selection and AdaBoost algorithm | |
| CN116755000A (en) | Transformer fault rapid identification device and method | |
| CN113496255B (en) | Power distribution network mixed observation point distribution method based on deep learning and decision tree driving | |
| CN116263814A (en) | Fault diagnosis method for oil immersed transformer | |
| CN113657726B (en) | Personnel risk analysis method based on random forest | |
| CN115153549A (en) | BP neural network-based man-machine interaction interface cognitive load prediction method | |
| CN115359857A (en) | Method for predicting PM2.5 concentration based on CART model and optimized RVFLNN model | |
| CN110852178A (en) | Piano music score difficulty identification method based on decision tree lifting | |
| CN116306226B (en) | Fuel cell performance degradation prediction method | |
| CN113239021B (en) | Data migration method for predicting residual life of similar products | |
| CN116680532A (en) | Transformer fault online diagnosis method for processing unbalanced small sample based on NNTR-SMOTE oversampling | |
| CN117473424A (en) | Transformer fault diagnosis method, system, equipment and medium based on random forest |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |