CN112465010A - DKPCA and KFDA-based transformer fault detection and classification method - Google Patents
DKPCA and KFDA-based transformer fault detection and classification method Download PDFInfo
- Publication number
- CN112465010A CN112465010A CN202011340810.3A CN202011340810A CN112465010A CN 112465010 A CN112465010 A CN 112465010A CN 202011340810 A CN202011340810 A CN 202011340810A CN 112465010 A CN112465010 A CN 112465010A
- Authority
- CN
- China
- Prior art keywords
- transformer
- fault
- data
- matrix
- formula
- 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
- 238000000034 method Methods 0.000 title claims abstract description 59
- 238000001514 detection method Methods 0.000 title claims abstract description 24
- 239000011159 matrix material Substances 0.000 claims abstract description 47
- 239000013598 vector Substances 0.000 claims abstract description 30
- 238000013507 mapping Methods 0.000 claims abstract description 25
- 230000003416 augmentation Effects 0.000 claims abstract description 11
- 238000005259 measurement Methods 0.000 claims abstract description 9
- 238000012549 training Methods 0.000 claims abstract description 8
- 238000012360 testing method Methods 0.000 claims description 16
- HYBBIBNJHNGZAN-UHFFFAOYSA-N furfural Chemical compound O=CC1=CC=CO1 HYBBIBNJHNGZAN-UHFFFAOYSA-N 0.000 claims description 6
- 229930195733 hydrocarbon Natural products 0.000 claims description 6
- 150000002430 hydrocarbons Chemical class 0.000 claims description 6
- 238000004804 winding Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- 239000004215 Carbon black (E152) Substances 0.000 claims description 4
- 238000009413 insulation Methods 0.000 claims description 4
- 230000000717 retained effect Effects 0.000 claims description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 4
- 235000009508 confectionery Nutrition 0.000 claims description 3
- 230000001186 cumulative effect Effects 0.000 claims description 3
- 238000013434 data augmentation Methods 0.000 claims description 3
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 239000006185 dispersion Substances 0.000 claims description 3
- 238000004519 manufacturing process Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 2
- 230000006870 function Effects 0.000 abstract description 11
- 238000003745 diagnosis Methods 0.000 abstract description 8
- 238000007781 pre-processing Methods 0.000 abstract 1
- 239000007789 gas Substances 0.000 description 16
- 238000009826 distribution Methods 0.000 description 2
- 238000010891 electric arc Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000000513 principal component analysis Methods 0.000 description 2
- 238000000926 separation method Methods 0.000 description 2
- 230000032683 aging Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000013021 overheating Methods 0.000 description 1
- 238000004540 process dynamic Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Business, Economics & Management (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Artificial Intelligence (AREA)
- Pure & Applied Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Public Health (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Primary Health Care (AREA)
- Computing Systems (AREA)
- Marketing (AREA)
- Human Resources & Organizations (AREA)
- General Health & Medical Sciences (AREA)
- Water Supply & Treatment (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- Housings And Mounting Of Transformers (AREA)
Abstract
The invention discloses a DKPCA and KFDA transformer fault detection and classification method, which comprises the steps of firstly collecting transformer operation data under the normal working condition of a transformer, and preprocessing the collected data; then establishing the correlation of the time-lag augmentation matrix reserved data, and extracting the characteristic information of the original data through the nonlinear mapping of the kernel function; establishing SPE and T through extracted score vectors2Statistics detecting occurrence of a fault; calculating the contributions RBC of each variable to the two statistics in the process, and determining the variable causing the fault; and finally, evaluating a fault variable by using KFDA to determine the fault type. According to the method, the DKPCA method is introduced into a transformer fault diagnosis method, so that the autocorrelation characteristic of the measured variable is kept, meanwhile, nonlinear characteristic information is accurately extracted, and the fault detection accuracy is improved; contribution system of introducing KFDA (Kalman Filter and data acquisition) to fault variableAnd the measurement is used for training and learning, so that the fault diagnosis precision of the transformer is improved.
Description
Technical Field
The invention belongs to the technical field of power transformer fault diagnosis of a power distribution network, relates to a transformer fault detection and classification method based on DKPCA and KFDA, and particularly relates to a low-voltage transformer fault detection and classification method of the power distribution network based on DKPCA and KFDA algorithms.
Background
Power transformers are used to convert different voltage levels in an electrical power system and are an important guarantee of reliability in the electrical power system. In order to ensure the reliability and safety of the power grid, maintenance personnel need to periodically acquire the operating state parameters of the transformer, and analyze the operating condition of the transformer through the parameters. However, manual maintenance depends on manual experience, the working efficiency is low, subjectivity is high, and the risk of transformer failure is increased. Therefore, under the background of artificial intelligence, the development of the intelligent transformer fault detection and diagnosis method has important practical significance.
At present, a method for analyzing Dissolved Gas (DGA) in transformer oil becomes a main method for judging the internal fault property of a transformer in a power system. The most common method in the DGA method is the three-ratio judgment method, so that the fault judgment accuracy is low, the existing fault detection method based on machine learning is difficult to mine the time sequence relation between measurement data, a nonlinear process is adopted before and after the fault occurs, and the reliability of the conventional machine learning method in transformer fault diagnosis needs to be improved.
Dynamic Kernel Principal Component Analysis (DKPCA) as a novel data-driven technique can well process dynamic nonlinear data. The basic idea is to obtain the dynamic characteristics among data by constructing a time lag matrix, and then map the data to a high-dimensional space by utilizing the nonlinear mapping of a kernel matrix, so that the time sequence correlation of the original data is kept, and meanwhile, the data which is difficult to separate is also separated in the high-dimensional space. Based on this, the feature information of the original data is extracted, and the time-series correlation features of the data are retained.
The Kernel Fisher Discriminant Analysis (KFDA) is used for determining the sample type according to the characteristics and the discriminant indexes of the sample data, and in order to solve the problem of nonlinear classification of fault contribution statistics, input data are subjected to kernel function nonlinear mapping, Fisher linear discriminant is used in a high-dimensional space, and the precision of fault diagnosis is improved.
In conclusion, the invention provides the transformer fault detection and classification method based on DKPCA and KFDA, which can effectively improve the transformer fault detection and classification precision and improve the operation reliability of the transformer.
Disclosure of Invention
In summary, in order to improve the fault classification accuracy of the power transformer, the invention provides a transformer fault detection and classification method based on DKPCA and KFDA. Firstly, extracting characteristic variables containing dynamic characteristics from massive redundant transformer state information through dynamic kernel principal component analysis, and utilizing T2Detecting whether the fault occurs by SPE statistics; and finally, calculating the contribution RBC corresponding to the fault statistics, and judging the fault category according to a KFDA classifier.
In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
a transformer fault detection and classification method based on DKPCA and KFDA comprises the following steps:
step 1, collecting transformer operation data, and performing normalization processing on the collected data by adopting a Z-score method to enable the mean value of the processed data to be zero and the variance to be 1;
step 2, superposing vectors at s moments before t moment after each variable to form a time lag augmentation matrix X(s) so as to reflect the dynamic relation between the variables;
step 3, finding the optimal hypersphere defined by the eigenvector v, establishing a dynamic kernel matrix to separate fault variables, reserving the maximum variance information according to the defined dynamic kernel principal element objective function, and establishing SPE and T through the extracted score vector2Statistics detecting occurrence of a fault;
step 4, respectively calculating each variable pair T in the process2Contribution of statisticsContribution to SPE statisticsDrawing a work contribution histogram, and judging a variable causing the fault according to the maximum contribution degree in the fault contribution graph;
and 5, determining the fault type of the transformer according to the KFDA judgment result.
Further, the step 1 further includes:
acquiring transformer operation data comprises acquiring dissolved gas data in transformer oil, transformer electrical test data and transformer insulating oil characteristic test data under the normal working condition of a transformer as a training sample set; taking the collected data as a training sample set x ∈ Rn×mWhere n is the number of measurement samples, each sample containing m observations; and carrying out normalization processing on the acquired data by adopting a Z-score method, wherein the formula is as follows:
after the mean value mu and the standard deviation sigma of the acquired data are calculated, the mean value of the processed data is 0 and the variance is 1 through the formula.
Further, the data of the dissolved gas in the transformer oil comprises H2、C2H2、CH4、C2H6、C2H4And the total hydrocarbon content; also included are the gas production rate of total hydrocarbons and CO2The ratio of the content of the stage CO gas; the transformer electrical test data comprise transformer dielectric loss factors, winding leakage current, insulation resistance and winding direct-current resistance phase difference; the transformer insulating oil characteristic test data comprise dielectric loss of insulating oil, water content in oil and furfural content.
Further, the step 2 further comprises:
after each normalized variable is superposed with the vectors of s times before t time, a time-lag augmentation matrix X(s) is established to reflect the dynamic relation between the variables, and a dynamic kernel matrix is established to separate fault variables
In the formula: x is the number oftRepresenting the samples taken at time t, s represents the time lag, and n is the number of samples.
Further, the step 3 further includes:
selecting kernel function as radial basis kernel function K ═ exp (— | | x-y | | | sweet wind path2/2σ2) Mapping the measurement sample to high-dimensional feature space, mapping a time-lag augmentation matrix X(s) to phi(s), and then establishing a data covariance matrix under the feature space
In the formula: phii(t: t-s) is the mapped dynamic data augmentation matrix;
carrying out eigenvector decomposition on the covariance matrix to find the optimal hypersphere defined by the eigenvector v
In the formula: alpha is alphaiIs a coefficient vector, v is a feature vector;
computing a dynamic kernel matrix
K=<Φi(t:t-s),Φj(t:t-s)>
The centralization treatment is carried out under the characteristic space, so thatCore matrix centralizationIn the formula: i isnIs an identity matrix;
objective function by dynamic kernel principal elements
In the formula: k is a dynamic kernel matrix;
solving for a score vectorWherein the number of principal elements retained is determined using a cumulative variance contribution ratio criterion (R)CPV>95%);
Establishing a Squared Prediction Error (SPE) and Hotelling's T based on the extracted score vector2Statistics determine control limits for faults:
T2=[t1,t2,…,tk]Λ-1[t1,t2,…,tk]T
in the formula: lambda-1And the inverse matrix of the diagonal matrix formed by the principal component eigenvalues is represented.
Further, the step 4 further includes:
In the formula: v. ofiIs the ith feature vector, xnewIs a new sample.
Further, the step 5 further includes:
dividing transformer faults into C types, wherein the fault variable number is N and is recorded as r ═ r1,r2,...,rCH, class i contains NiOne sample is recorded asAfter nonlinear mapping, r → phi (r)i) The mean value after mapping the ith sample in the high-dimensional space is miThe mean value after mapping all samples is m;
calculating intra-kernel dispersion SWAnd degree of interspinous divergence SB
In the formula: m isiThe mean value of the i-th sample after mapping, and m is the mean value of all samples after mapping;
establishing Fisher criterion in high-dimensional space
In the formula: w is aφIs any non-zero column vector;
and judging the fault type by adopting the Mahalanobis distance.
Compared with the prior art, the method has the advantages that:
1) the invention captures the dynamic correlation of the data along with the time change by constructing the time-delay augmentation matrix, and the extracted dynamic characteristics can comprehensively analyze the data potential information.
2) And a better data separation effect is obtained by utilizing the nonlinear mapping of the kernel function.
3) And identifying fault variables according to the contribution RBC corresponding to the fault statistics, so that the fault reason can be clarified.
4) And further classifying fault variables through KFDA (Kalman Filter/data acquisition) to improve the fault classification precision.
Drawings
Fig. 1 is a flow chart of a DKPCA-FDA based transformer fault detection and diagnosis method of the present invention.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
The invention provides a DKPCA and KFDA-based transformer fault detection and diagnosis method, the flow chart of which is shown in FIG. 1 and is specifically implemented according to the following steps:
step 1: acquiring transformer operation data, and performing normalization processing on the acquired data by adopting a Z-score method to ensure that the mean value of the processed data is zero and the variance is 1;
further, the step 1 also comprises the following implementation process of collecting dissolved gas data in oil under the normal working condition of the transformer, electrical test data of the transformer and characteristic test data of insulating oil as a training sample set;
further, the step 1 should also include the following implementation processes: the data of dissolved gas in transformer oil comprises H2、C2H2、CH4、C2H6、C2H4The content of various gases such as total hydrocarbon, the gas production rate of the total hydrocarbon and CO2The ratio of the content of the stage CO gas; the electrical test data of the transformer comprise dielectric loss factors of the transformer, winding leakage current,The phase difference of the insulation resistance and the winding direct-current resistance; the insulating oil characteristic test data comprise dielectric loss of the insulating oil, water content in the oil and content of furfural.
Further, the step 1 should also include the following implementation processes: dissolved gas data in transformer oil, transformer electrical test data and insulating oil characteristic test data acquired under normal working condition are used as a training sample set x ∈ Rn×mAnd the real-time collected data of dissolved gas in the transformer oil, the electrical test data of the transformer and the characteristic test data of the insulating oil are used as a training sample set x ∈ Rn×mWhere n is the number of measurement samples, each sample containing m observations;
further, the step 1 should also include the following implementation processes: and (3) normalizing the measured data by adopting a Z-score method, wherein the mean value of the processed data is 0, and the variance is 1.
After calculating the mean μ and standard deviation σ of the measurement data, the mean of the processed data was made 0 and the variance was made 1 by equation (1).
Step 2: superposing vectors s times before t time after each variable to form a time lag augmentation matrix X(s) so as to reflect the dynamic relation among the variables and establish a dynamic kernel matrix separation fault variable;
further, the step 2 should also include the following implementation processes: establishing a time-lag augmentation matrix to reflect the dynamic relationship between the data:
further, the step 2 should also include the following implementation processes: determining the time s of the time-lag matrix through a recurrence formula, wherein the specific method comprises the following steps:
1.s=0;
2. executing PCA algorithm and calculating all principal elements;
3.j=n×(s+1),r(s)=0;
if jth principal element is linear
5.Then j=j-1,r(s)=r(s)+1;
7.End if
8.While rnew(s)>0
9.Do s=s+1,go to step 2;
10.Else output s;
11.End
And step 3: finding out the optimal hypersphere defined by the eigenvector v, establishing a dynamic kernel matrix to separate fault variables, reserving the maximum variance information according to the defined dynamic kernel principal element objective function, and establishing SPE and T through the extracted score vector2Statistics detecting occurrence of a fault;
further, the step 3 should also include the following implementation processes: selecting kernel function as radial basis kernel function K ═ exp (— | | x-y | | | sweet wind path2/2σ2) Mapping the measurement sample to high-dimensional feature space, mapping a time-lag augmentation matrix X(s) to phi(s), and then establishing a data covariance matrix under the feature space
Carrying out eigenvector decomposition on the covariance matrix to find the optimal hypersphere defined by the eigenvector v
Further, the step 3 should further include the following processes: computing a dynamic kernel matrix
K=<Φi(t:t-s),Φj(t:t-s)> (5)
The centralization treatment is carried out under the characteristic space, so thatCore matrix centralization
Further, the step 3 should further include the following processes: objective function by dynamic kernel principal elements
Solving for a score vectorWherein the number of principal elements retained is determined using a cumulative variance contribution ratio criterion (R)CPV>95%)。
Further, the step 3 should further include the following processes: establishing a Squared Prediction Error (SPE) and Hotelling's T based on the extracted score vector2Statistics determine control limits for faults:
T2=[t1,t2,…,tk]Λ-1[t1,t2,…,tk]T (7)
and 4, step 4: respectively calculating each variable pair T in the process2Contribution of statisticsContribution to SPE statisticsDrawing a work contribution histogram, and judging a variable causing the fault according to the maximum contribution degree in the fault contribution graph;
further, the step 4 should further include the following processes: t is2Contribution of statisticsAnd contribution of SPE statisticsRespectively is as follows;
and 5: determining the fault type of the transformer according to a KFDA (Kalman Filter/direct Current) judgment result;
further, the step 5 should further include the following processes: dividing transformer faults into C types, wherein the fault variable number is N and is recorded as r ═ r1,r2,...,rCH, class i contains NiOne sample is recorded asAfter nonlinear mapping, r → phi ri) The mean value after mapping the ith sample in the high-dimensional space is miThe mean value after mapping all samples is m;
further, the step 5 should further include the following processes: calculating intra-kernel dispersion SWAnd degree of interspinous divergence SB;
Further, the step 5 should further include the following processes: establishing Fisher criterion in a high-dimensional space:
further, the step 5 should further include the following processes: judging the fault type by adopting the Mahalanobis distance;
further, the step 5 should further include the following processes: the failure categories and corresponding gas compositions are shown in table 1.
TABLE 1 Fault types and corresponding gas composition tables
Main gas component | Secondary gas component | Type of failure |
H2 | / | The water or oil containing bubbles |
CO、CO2 | / | Aging of oil |
CH4 | C2H4 | Oil superheating |
CH4、C2H4、CO、CO2 | H2、C2H6 | Severe overheating of oil and paper |
H2、CH4 | C2H2、C2H6 | Oiled paper insulation partial discharge |
CH4、C2H2、H2 | / | Spark discharge in oil |
H2、C2H2、CH4 | C2H4、C2H6 | Electric arc in oil |
H2、C2H2、CO、CO2 | CH4、C2H6、C2H4 | Electric arc in oil and paper |
In the above formulae (1) to (15), xtRepresenting the samples taken at time t, s represents the time lag, n is the number of samples, phii(t: t-s) is the mapped dynamic data augmentation matrix, αiIs a coefficient vector, v is a feature vector, InIs an identity matrix, K is a dynamic kernel matrix, Λ-1An inverse matrix, v, representing a diagonal matrix of principal component eigenvaluesiIs the ith feature vector, xnewIs a new sample of the sample, and,representing the ith class of jth sample vectors, m, in a feature spaceiIs the mean value of the i-th sample after mapping, m is the mean value of all samples after mapping, wφIs any non-zero column vector.
Claims (10)
1. A transformer fault detection and classification method based on DKPCA and KFDA is characterized in that:
step 1, collecting transformer operation data, and performing normalization processing on the collected data by adopting a Z-score method to enable the mean value of the processed data to be zero and the variance to be 1;
step 2, superposing vectors at s moments before t moment after each variable to form a time lag augmentation matrix X(s) so as to reflect the dynamic relation between the variables;
step 3, finding the optimal hypersphere defined by the eigenvector v, establishing a dynamic kernel matrix to separate fault variables, reserving the maximum variance information according to the defined dynamic kernel principal element objective function, and establishing SPE and T through the extracted score vector2Statistics detecting occurrence of a fault;
step 4, respectively calculating each variable pair T in the process2Contribution of statisticsContribution to SPE statisticsDrawing a work contribution histogram, and judging a variable causing the fault according to the maximum contribution degree in the fault contribution graph;
and 5, determining the fault type of the transformer according to the KFDA judgment result.
2. The transformer fault detection and classification method according to claim 1, wherein the step 1 further comprises:
acquiring transformer operation data comprises acquiring dissolved gas data in transformer oil, transformer electrical test data and transformer insulating oil characteristic test data under the normal working condition of a transformer as a training sample set;
taking the collected data as a training sample set x ∈ Rn×mWhere n is the number of measurement samples, each sample containing m observations;
and carrying out normalization processing on the acquired data by adopting a Z-score method, wherein the formula is as follows:
after the mean value mu and the standard deviation sigma of the acquired data are calculated, the mean value of the processed data is 0 and the variance is 1 through the formula.
3. The transformer fault detection and classification method according to claim 2, wherein the step 2 further comprises:
after each normalized variable is superposed with the vectors of s times before t time, a time-lag augmentation matrix X(s) is established to reflect the dynamic relation between the variables, and a dynamic kernel matrix is established to separate fault variables
In the formula: x is the number oftRepresenting the samples taken at time t, s represents the time lag, and n is the number of samples.
4. The transformer fault detection and classification method according to claim 3, wherein the step 3 further comprises:
selecting kernel function as radial basis kernel function K ═ exp (— | | x-y | | | sweet wind path2/2σ2) Mapping the measurement samples to a high-dimensional feature space, and mapping a time-lag augmentation matrix X(s) intoPhi(s), and then establishing a data covariance matrix under the feature space
In the formula: phii(t: t-s) is the mapped dynamic data augmentation matrix;
carrying out eigenvector decomposition on the covariance matrix to find the optimal hypersphere defined by the eigenvector v
In the formula: alpha is alphaiIs a coefficient vector, v is a feature vector;
computing a dynamic kernel matrix
K=<Φi(t:t-s),Φj(t:t-s)>
The centralization treatment is carried out under the characteristic space, so thatCore matrix centralizationIn the formula: i isnIs an identity matrix;
objective function by dynamic kernel principal elements
s.t.αTKα=1
In the formula: k is a dynamic kernel matrix;
solving for a score vectorWherein the number of principal elements retained is determined using a cumulative variance contribution ratio criterion (R)CPV>95%);
Establishing a Squared Prediction Error (SPE) and Hotelling's T based on the extracted score vector2Statistics determine control limits for faults:
T2=[t1,t2,…,tk]Λ-1[t1,t2,…,tk]T
in the formula: lambda-1And the inverse matrix of the diagonal matrix formed by the principal component eigenvalues is represented.
6. The transformer fault detection and classification method according to claim 5, wherein the step 5 further comprises:
classifying transformer faults into C-type faultsThe variable number is N, and is recorded as r ═ r1,r2,...,rCH, class i contains NiOne sample is denoted as ri={ri 1,ri 2,...,ri NiR → phi (r) after nonlinear mappingi) The mean value after mapping the ith sample in the high-dimensional space is miThe mean value after mapping all samples is m;
in the formula: phi (r)i j)(i=1,...,m;j=1,...,ni) Representing the ith class jth sampling vector in the feature space;
calculating intra-kernel dispersion SWAnd degree of interspinous divergence SB
In the formula: m isiThe mean value of the i-th sample after mapping, and m is the mean value of all samples after mapping;
establishing Fisher criterion in high-dimensional space
In the formula: w is aφIs any non-zero column vector;
and judging the fault type by adopting the Mahalanobis distance.
7. The transformer fault detection and classification method according to any one of claims 2 to 6, characterized in that:
the data of the dissolved gas in the transformer oil comprises H2、C2H2、CH4、C2H6、C2H4And the total hydrocarbon content.
8. The transformer fault detection and classification method according to any one of claims 2 to 6, characterized in that:
the transformer electrical test data comprise transformer dielectric loss factors, winding leakage current, insulation resistance and winding direct-current resistance phase difference.
9. The transformer fault detection and classification method according to any one of claims 2 to 6, characterized in that:
the transformer insulating oil characteristic test data comprise dielectric loss of insulating oil, water content in oil and furfural content.
10. The transformer fault detection and classification method according to claim 7, characterized in that:
also included are the gas production rate of total hydrocarbons and CO2Ratio of the gas content of the stage CO.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011340810.3A CN112465010A (en) | 2020-11-25 | 2020-11-25 | DKPCA and KFDA-based transformer fault detection and classification method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011340810.3A CN112465010A (en) | 2020-11-25 | 2020-11-25 | DKPCA and KFDA-based transformer fault detection and classification method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112465010A true CN112465010A (en) | 2021-03-09 |
Family
ID=74808326
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011340810.3A Pending CN112465010A (en) | 2020-11-25 | 2020-11-25 | DKPCA and KFDA-based transformer fault detection and classification method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112465010A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113469257A (en) * | 2021-07-07 | 2021-10-01 | 云南大学 | Distribution transformer fault detection method and system |
CN113657556A (en) * | 2021-09-23 | 2021-11-16 | 华北电力大学 | Gas turbine inlet guide vane system fault diagnosis method based on multivariate statistical analysis |
CN117290670A (en) * | 2023-11-27 | 2023-12-26 | 南京中鑫智电科技有限公司 | Transformer bushing insulation state estimation method based on enhancement filter algorithm |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103412557A (en) * | 2013-07-17 | 2013-11-27 | 浙江工业大学 | Industrial fault detection and diagnostic method suitable for nonlinear process on-line monitoring |
US20150127277A1 (en) * | 2013-11-06 | 2015-05-07 | Electric Power Research Institute, Inc. | System and method for assessing power transformers |
CN107608335A (en) * | 2017-09-14 | 2018-01-19 | 山东科技大学 | A kind of UAV Flight Control System fault detect and the data-driven method of fault reconstruction |
CN108062565A (en) * | 2017-12-12 | 2018-05-22 | 重庆科技学院 | Double pivots-dynamic kernel principal component analysis method for diagnosing faults based on chemical industry TE processes |
CN111811894A (en) * | 2020-07-11 | 2020-10-23 | 国网辽宁省电力有限公司电力科学研究院 | Gas on-line detection method and device in power transformer gas relay |
-
2020
- 2020-11-25 CN CN202011340810.3A patent/CN112465010A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103412557A (en) * | 2013-07-17 | 2013-11-27 | 浙江工业大学 | Industrial fault detection and diagnostic method suitable for nonlinear process on-line monitoring |
US20150127277A1 (en) * | 2013-11-06 | 2015-05-07 | Electric Power Research Institute, Inc. | System and method for assessing power transformers |
CN107608335A (en) * | 2017-09-14 | 2018-01-19 | 山东科技大学 | A kind of UAV Flight Control System fault detect and the data-driven method of fault reconstruction |
CN108062565A (en) * | 2017-12-12 | 2018-05-22 | 重庆科技学院 | Double pivots-dynamic kernel principal component analysis method for diagnosing faults based on chemical industry TE processes |
CN111811894A (en) * | 2020-07-11 | 2020-10-23 | 国网辽宁省电力有限公司电力科学研究院 | Gas on-line detection method and device in power transformer gas relay |
Non-Patent Citations (15)
Title |
---|
JUN LIN 等: "Fault diagnosis of transformer based on KPCA and Elman neural network", 《2018 IEEE INTERNATIONAL CONFERENCE ON INDUSTRIAL TECHNOLOGY(ICIT)》 * |
KAN TANG 等: "Power Transformer Anomaly Detection Based on Adaptive Kernel Fuzzy C-Means Clustering and Kernel Principal Component Analysis", 《2018 AUSTRALIAN & NEW ZEALAND CONTROL CONFERENCE(ANZCC)》 * |
SANG WOOK CHOI 等: "Nonlinear dynamic process monitoring based on dynamic kernel PCA", 《CHEMICAL ENGINEERING SCIENCE》 * |
ZHANG XI 等: "Nonlinear online process monitoring and fault diagnosis of condenser based on kernel PCA plus FDA", 《JOURNAL OF SOUTHEAST UNIVERSITY(ENGLISH EDITION)》 * |
唐勇波 等: "重构贡献在变压器故障诊断中的应用", 《计算机工程》 * |
姜飞宇 等: "变压器绝缘套管油中溶解气体分析方法与问题", 《中国测试》 * |
宋玉琴 等: "改进KFDA分类器在电力变压器故障诊断中的应用", 《传感器与微系统》 * |
张起 等: "基于动态核主元分析的电缆接头故障检测方法", 《大连理工大学学报》 * |
李涵深: "基于数据驱动的变压器故障诊断与智能预警", 《中国优秀硕士学位论文全文数据库 (工程科技Ⅱ辑)》 * |
李霖: "基于BP神经网络的变压器故障诊断方法研究", 《中国优秀硕士学位论文全文数据库 (工程科技Ⅱ辑)》 * |
杜景琦 等: "电厂发电设备故障诊断方法综述", 《云南电力技术》 * |
梅红争: "基于油中气体分析的电力变压器故障诊断研究", 《中国优秀硕士学位论文全文数据库 (工程科技Ⅰ辑)》 * |
王涛: "面向电力变压器的运行状态综合评价", 《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅱ辑》 * |
赵朋: "变压器油中特征气体监测技术及应用的研究", 《中国优秀硕士学位论文全文数据库 (工程科技Ⅱ辑)》 * |
马立玲 等: "一种基于改进核Fisher的故障诊断方法", 《化工学报》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113469257A (en) * | 2021-07-07 | 2021-10-01 | 云南大学 | Distribution transformer fault detection method and system |
CN113657556A (en) * | 2021-09-23 | 2021-11-16 | 华北电力大学 | Gas turbine inlet guide vane system fault diagnosis method based on multivariate statistical analysis |
CN113657556B (en) * | 2021-09-23 | 2023-12-26 | 华北电力大学 | Gas turbine inlet guide vane system fault diagnosis method based on multivariate statistical analysis |
CN117290670A (en) * | 2023-11-27 | 2023-12-26 | 南京中鑫智电科技有限公司 | Transformer bushing insulation state estimation method based on enhancement filter algorithm |
CN117290670B (en) * | 2023-11-27 | 2024-01-26 | 南京中鑫智电科技有限公司 | Transformer bushing insulation state estimation method based on enhancement filter algorithm |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112465010A (en) | DKPCA and KFDA-based transformer fault detection and classification method | |
CN109685138B (en) | XLPE power cable partial discharge type identification method | |
CN108803520B (en) | Dynamic process monitoring method based on variable nonlinear autocorrelation rejection | |
Abu-Elanien et al. | Determination of transformer health condition using artificial neural networks | |
CN107132310B (en) | Transformer equipment health status method of discrimination based on gauss hybrid models | |
CN104020401B (en) | The appraisal procedure of transformer insulated heat ageing state based on cloud models theory | |
CN107862114A (en) | Small echo PSO SVM Diagnosis Method of Transformer Faults based on three ratio feature amounts | |
CN110262450B (en) | Fault prediction method for cooperative analysis of multiple fault characteristics of steam turbine | |
CN107491783B (en) | Transformer fault type classification method based on improved density peak value clustering algorithm | |
Chen et al. | Fault diagnosis in chemical process based on self-organizing map integrated with fisher discriminant analysis | |
CN112183590A (en) | Transformer fault diagnosis method based on Oneclass SVM algorithm | |
CN113782113B (en) | Method for identifying gas fault in transformer oil based on deep residual error network | |
CN109184821B (en) | Online monitoring method for closed-loop information analysis of steam turbine of large-scale generator set | |
CN114595883A (en) | Oil-immersed transformer residual life personalized dynamic prediction method based on meta-learning | |
CN111275204A (en) | Transformer state identification method based on hybrid sampling and ensemble learning | |
CN114116832A (en) | Power distribution network abnormity identification method based on data driving | |
CN114091549A (en) | Equipment fault diagnosis method based on deep residual error network | |
Xinyi et al. | Fault diagnosis in chemical process based on self-organizing map integrated with fisher discriminant analysis | |
Aciu et al. | Determination of the condition of solid insulation in high-power transformers based on 2-furfuraldehyde and methanol markers using neural networks | |
Sarma et al. | A long short-term memory based prediction model for transformer fault diagnosis using dissolved gas analysis with digital twin technology | |
CN114019365A (en) | On-load tap-changer fault diagnosis method based on gas detection technology in oil | |
CN108508866B (en) | Million kilowatt unit fault variable identification method based on sparse relative discriminant analysis | |
Dadzie et al. | Transformer incipient fault detection technique based on neural network | |
Wang et al. | Raman diagnosis method for thermal aging of insulating paper based on AE-LDA and naive Bayes | |
CN111626559A (en) | Main factor analysis method-based medium-voltage distribution network line loss key characteristic index extraction method and system |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210309 |
|
RJ01 | Rejection of invention patent application after publication |