CN114528921B - Transformer fault diagnosis method based on LOF algorithm and mixed sampling - Google Patents
Transformer fault diagnosis method based on LOF algorithm and mixed sampling Download PDFInfo
- Publication number
- CN114528921B CN114528921B CN202210064482.1A CN202210064482A CN114528921B CN 114528921 B CN114528921 B CN 114528921B CN 202210064482 A CN202210064482 A CN 202210064482A CN 114528921 B CN114528921 B CN 114528921B
- Authority
- CN
- China
- Prior art keywords
- point
- sample
- data
- fault
- samples
- 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.)
- Active
Links
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 31
- 238000003745 diagnosis Methods 0.000 title claims abstract description 23
- 238000005070 sampling Methods 0.000 title claims abstract description 22
- 238000000034 method Methods 0.000 title claims abstract description 17
- 230000002159 abnormal effect Effects 0.000 claims description 11
- 238000012549 training Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000007781 pre-processing Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 4
- 238000007637 random forest analysis Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims 2
- 239000007789 gas Substances 0.000 description 8
- 238000012423 maintenance Methods 0.000 description 5
- 230000002194 synthesizing effect Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 2
- 230000032683 aging Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000009413 insulation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/50—Testing of electric apparatus, lines, cables or components for short-circuits, continuity, leakage current or incorrect line connections
- G01R31/62—Testing of transformers
-
- 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
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- 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
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Power Engineering (AREA)
- Probability & Statistics with Applications (AREA)
- Testing Relating To Insulation (AREA)
Abstract
The invention discloses a transformer fault diagnosis method based on an LOF algorithm and a mixed sampling, which solves the problem of unbalanced data by adopting single oversampling or undersampling in the traditional data balancing method, but the single sampling is very likely to cause important data loss, and a plurality of useless noise data are increased at the same time so as to influence the accuracy of a subsequent fault diagnosis model.
Description
Technical Field
The invention relates to a transformer fault diagnosis method based on an LOF algorithm and mixed sampling, and belongs to the technical field of transformer fault diagnosis.
Background
Along with the continuous development of national economy and continuous improvement of living standard, the demand of society for electric power is increasingly larger, so that in order not to influence the normal electricity consumption of residents, the stable operation of an electric power system becomes an important component part in the daily work of operation and maintenance personnel, a power transformer is used as one of key electric equipment for electric energy conversion and energy output of the electric power system, the operation state of the power transformer has very direct influence on the normal and stable operation of the electric power system, so that the maintenance personnel is required to pay high attention to the maintenance personnel to keep the good operation state of the transformer, and once the transformer fails, the maintenance personnel is required to rapidly and accurately position the failure position and maintain the failure position, so that the follow-up power supply reliability is ensured. Therefore, the research of the fault diagnosis method of the power transformer has important significance for guiding maintenance work. The traditional analysis method of the dissolved gas in the oil can effectively reflect the early failure of the transformer, is widely applied to practical scenes, and still has the problems of low diagnosis accuracy and the like caused by unbalanced proportion of failure data and normal data. Therefore, the invention provides a transformer fault diagnosis method based on an LOF algorithm and mixed sampling.
Disclosure of Invention
The invention provides a transformer fault diagnosis method based on an LOF algorithm and mixed sampling, which is used for solving the problem of unbalance of training data before training a model, reducing the generation of noise data after sampling and guaranteeing the accuracy of subsequent transformer fault diagnosis.
In order to achieve the purpose of the invention, the invention adopts the following technical scheme:
The invention discloses a transformer fault diagnosis method based on an LOF algorithm and mixed sampling, which comprises the following steps:
S1, collecting data samples. Failure data D 1,D2,T1,T2 and normal data NC in the dissolved gas in the oil are collected, resulting in an initial unbalanced data set.
S2, data preprocessing. Normalizing the unbalanced data set to reduce the numerical range of the data set:
Further, the specific implementation of step 2:
The initial dataset is brought into a normalized formula calculation, which is as follows:
Where z ij represents the original data, Representing the normalized data; i represents the number of sample data; j represents a characteristic dimension number, and 1 to 5 represent five gas type data, respectively.
S3, in order to avoid influencing the diagnosis result of the subsequent model, removing abnormal values in the data set, and respectively removing abnormal values in the fault data sample set and the normal data sample set by adopting an LOF algorithm;
Further, the specific implementation of the step 3 is as follows:
(1) Calculating sample point spacing: for n samples to be tested, the distance between any two sample points is calculated by adopting the mahalanobis distance, and the mahalanobis distance formula is as follows:
The reachable distance between the sample point O to be measured and the kth local cluster point is calculated, and the formula is as follows:
distk(O,P)＝MAX{distk(O),dist(O,P)}
the right side of the equation shows that the result takes the maximum of the distance of the point from the point O and the distance of the kth cluster point P from the point O.
(2) Calculating local reachable density lrd (O) of the sample point O to be measured:
Where N k (O) represents the kth distance field of the point O, that is, the kth distance of the point O and all points within the kth distance, including the kth distance, |n k (O) | represents the number of the kth field points of the point O, and d k (O, P) represents the distances between the sample point O to be measured and the kth cluster point P.
(3) Calculating local outlier factors LOF, wherein the LOF value reflects the difference between the reachable density of a certain point and the density of points in other fields and is used as a basis for judging whether the point O to be measured is an abnormal value or not:
where lrd (P) and lrd (O) represent local reachable densities of point P and point O, respectively, if the LOF value is closer to 1, the reachable density of point O is similar to the local density of point O, if the LOF value is smaller than 1, the reachable density of point O is higher than the local density of point O, and if the LOF value is larger than 1, the reachable density of point O is lower than the local density of point O, which is regarded as an outlier.
S4, expanding few types of fault sample data. Oversampling is carried out by adopting an SMOTE algorithm;
further, the specific implementation of step 4:
(1) Setting the number of fault samples as T, calculating the sample interval by using the Euclidean distance, k neighbor samples x i(near) are obtained from the failure samples, near e1, once again, k }.
(2) Setting a sampling multiplying power N according to an unbalanced proportion, for each original fault sample x i, randomly selecting one sample x i(nn) from k neighbor samples of the original fault sample x i, and synthesizing the original fault sample and the neighbor fault samples into a new fault sample according to the following formula:
xnew＝xi+rand(0,1)*|xi-xi(nn)|
(3) And repeating the two steps for N times, and synthesizing N new fault samples x i(new), new epsilon { 1..once, N }, wherein the sample set at the moment is the new fault sample set.
S5, extracting a plurality of types of normal sample data. Undersampling a normal sample by adopting CRIUS algorithm;
further, the specific implementation of step 5:
(1) Firstly, clustering a data set into a plurality of clusters, and dividing a normal sample set into K clusters by using a K-means algorithm, wherein the K-means algorithm comprises the following steps:
a) Firstly, K normal points are selected as initial centroids;
b) Secondly, calculating Euclidean distance from a point of a centroid neighbor to the centroid, and assigning each neighbor point to the nearest centroid to form K clusters;
c) Updating the mass center of each cluster according to the divided clusters; finally, the first two steps are repeated until the mass center of the cluster group does not move.
(2) Randomly selecting M normal samples in each cluster group which has been clustered to form a subset X +, wherein
(3) The information potential V (X) of each X + is then calculated as followsRepresenting a gaussian kernel function:
(4) The crossover information potential V (X, Y) between each normal sample is calculated as follows:
(5) And constructing a cost function and solving the minimum value of the cost function, and finding a sample subset X + capable of minimizing the cost function, wherein the sample subset is M samples representing normal samples after undersampling. The formula is as follows:
J(X+)＝H2(X+)+λDCS(X+,X+)
Where H 2(X+) is Renyi quadratic entropy for measuring normal redundancy and D CS(X+,X+) is CS divergence for processing distortion in undersampled data, the variables having the following relationship:
DCS(X,Y)＝2H2(X,Y)-H2(X)-H2(Y)
the cost function is optimized according to the relation to obtain the following formula:
Wherein the parameter lambda epsilon R + is mainly used to control the distortion level in the subsampled data.
(6) The samples in subset X + are updated and the iterative formula is as follows:
Where X +k∈X+ (k e {1,2,., M }) refers to the samples in subset X +, Representing the gaussian kernel function,T.epsilon.N means that t iterations have been performed.
(7) Repeating the steps (3) to (6) until the result converges, thereby obtaining M most representative normal samples.
S6, training a model. And integrating the normal sample after the secondary sampling with the fault sample to obtain a new balance sample set, inputting the new balance sample set into a random forest model for training, thereby establishing a fault diagnosis model, and classifying the five types of low-energy discharge (D 1), high-energy discharge (D 2), medium-low temperature overheat (T 1), high-temperature overheat (T 2) and normal data (NC) of the transformer.
The beneficial effects of the invention are as follows:
1. according to the invention, firstly, the LOF algorithm is utilized to remove abnormal values such as noise and the like in a data preprocessing stage, and then the combined CRIUS algorithm and the SMOTE algorithm are mixed and sampled to perform data balancing, so that the advantages of the two algorithms are fully exerted, noise data generation is reduced while the data balancing is processed, and data with important information are reserved, so that the accuracy of a diagnosis model is improved, and a better guiding effect is provided for timely finding transformer faults. .
2. Conventional data balancing methods generally use a single over-sampling or under-sampling to solve the data imbalance problem, but the single sampling is highly likely to cause important data loss, and meanwhile, many useless noise data are added, so that the accuracy of a subsequent fault diagnosis model is affected. According to the invention, the LOF algorithm is adopted to remove noise or abnormal value before data balancing, and then CRIUS algorithm and SMOTE algorithm are utilized to carry out mixed sampling, so that the influence caused by the traditional method is overcome, and the accuracy of the diagnosis model is improved.
Drawings
Fig. 1: transformer fault diagnosis method flow chart based on LOF algorithm and mixed sampling
Detailed Description
The invention is further described below with reference to the drawings and examples.
In this embodiment, a transformer fault diagnosis method based on LOF algorithm and mixed sampling, as shown in fig. one, includes:
S1, generating gases with different components by internal chemical substances due to insulation aging, thermal faults, electric faults and the like in the transformer, wherein the gases comprise characteristic gases such as H 2、CH4、C2H6、C2H4、C2H2, and collecting fault data and normal data of the volume content of dissolved gases in oil, wherein the fault data and the normal data comprise low-energy discharge D 1, high-energy discharge D 2, medium-low temperature overheat T 1, high-temperature overheat T 2 and normal data NC, and the fault data and the normal data are used as an initial data set z ij.
S2, data preprocessing.
Normalizing the unbalanced data set to reduce the numerical range of the data set;
Further, the specific implementation of step 2:
The initial dataset is brought into a normalized formula calculation, which is as follows:
Where z ij represents the original data, Representing the normalized data; i represents the number of sample data; j represents a characteristic dimension number, and 1 to 5 represent five gas type data, respectively.
S3, removing abnormal values, such as noise in the data set, which is one of important factors for reducing the training effect of the model, and adopting LOF (Local Outlier Factor) algorithm to remove abnormal values in the fault data and normal data sample set respectively in order to avoid influencing the diagnosis result of the subsequent model;
Further, the specific implementation of the step 3 is as follows:
(1) Calculating sample point spacing: for n samples to be tested, the distance between any two sample points is calculated by adopting the mahalanobis distance, and the mahalanobis distance formula is as follows:
The reachable distance between the sample point O to be measured and the kth local cluster point is calculated, and the formula is as follows:
distk(O,P)＝MAX{distk(O),dist(O,P)}
The right side of the equation shows that the result takes the maximum value of the distance dist k (O) from this point to point O and the distance dist (O, P) from the kth cluster point P to point O.
(2) Calculating local reachable density lrd (O) of the sample point O to be measured:
Where N k (O) represents the kth distance field of the point O, that is, the kth distance of the point O and all points within the kth distance, including the kth distance, |n k (O) | represents the number of the kth field points of the point O, and d k (O, P) represents the distances between the sample point O to be measured and the kth cluster point P.
(3) Calculating local outlier factors LOF, wherein the LOF value reflects the difference between the reachable density of a certain point and the density of points in other fields and is used as a basis for judging whether the point O to be measured is an abnormal value or not:
where lrd (P) and lrd (O) represent local reachable densities of point P and point O, respectively, if the LOF value is closer to 1, the reachable density of point O is similar to the local density of point O, if the LOF value is smaller than 1, the reachable density of point O is higher than the local density of point O, and if the LOF value is larger than 1, the reachable density of point O is lower than the local density of point O, which is regarded as an outlier.
S4, expanding fault sample data. Oversampling is carried out by adopting an SMOTE algorithm;
further, the specific implementation of step 4:
(1) Setting the number of fault samples as T, calculating the sample interval by using the Euclidean distance, k neighbor samples x i(near) are obtained from the failure samples, near e1, once again, k }.
(2) Setting a sampling multiplying power N according to an unbalanced proportion, for each original fault sample x i, randomly selecting one sample x i(nn) from k neighbor samples of the original fault sample x i, and synthesizing the original fault sample and the neighbor fault samples into a new fault sample according to the following formula:
xnew＝xi+rand(0,1)*|xi-xi(nn)|
(3) And repeating the two steps for N times, and synthesizing N new fault samples x i(new), new epsilon { 1..once, N }, wherein the samples at the moment are the fault samples.
S5, extracting normal sample data. Undersampling a normal sample by adopting CRIUS algorithm;
further, the specific implementation of step 5:
(1) Firstly, clustering a data set into a plurality of clusters, and dividing a normal sample into K clusters by using a K-means algorithm, wherein the K-means algorithm comprises the following steps:
a) Firstly, K normal points are selected as initial centroids;
b) Secondly, calculating Euclidean distance from a point of a centroid neighbor to the centroid, and assigning each neighbor point to the nearest centroid to form K clusters;
c) Updating the mass center of each cluster according to the divided clusters; finally, the first two steps are repeated until the mass center of the cluster group does not move.
(2) Randomly selecting M normal samples in each normal cluster group which has been clustered to form a subset X +, wherein
(3) The information potential V (X) for each X + is then calculated, as shown below,Representing a gaussian kernel function:
(4) The crossover information potential V (X, Y) between each normal sample is calculated as follows:
(5) And constructing and solving a cost function minimum value, and finding a sample subset X + capable of minimizing the cost function, wherein the sample subset is M samples representing normal samples after undersampling. The formula is as follows:
J(X+)＝H2(X+)+λDCS(X+,X+)
Wherein H 2(X+) is Renyi quadratic entropy for measuring normal redundancy, D CS(X+,X+) is CS divergence for processing distortion in undersampled data, and lambda E R + is used for controlling the distortion degree of the sampled data, the variables have the following relation:
DCS(X,Y)＝2H2(X,Y)-H2(X)-H2(Y)
the cost function is optimized according to the relation to obtain the following formula:
Wherein the parameter lambda epsilon R + is mainly used to control the distortion level in the subsampled data.
(6) The samples in subset X + are updated and the iterative formula is as follows:
Where X +k∈X+ (k e {1,2,., M }) refers to the samples in subset X +, Representing the gaussian kernel function,T.epsilon.N means that t iterations have been performed.
(7) Repeating the steps (3) to (6) until the result converges, thereby obtaining M most representative normal samples.
S6, training a model. And integrating the normal sample after the secondary sampling with the fault sample to obtain a new balance sample set, inputting the new balance sample set into a random forest model for training, thereby establishing a fault diagnosis model, and classifying the five types of low-energy discharge (D 1), high-energy discharge (D 2), medium-low temperature overheat (T 1), high-temperature overheat (T 2) and normal data (NC) of the transformer.
The above list of detailed descriptions is only specific to practical embodiments of the present invention, and they are not intended to limit the scope of the present invention, and all equivalent manners or modifications that do not depart from the technical scope of the present invention should be included in the scope of the present invention.
Claims (1)
1. The transformer fault diagnosis method based on the LOF algorithm and the mixed sampling is characterized by comprising the following steps of:
S1, collecting a data sample; collecting fault data D 1,D2,T1,T2 and normal data NC in dissolved gas in oil to obtain an initial unbalanced data set;
s2, preprocessing data;
S3, respectively removing abnormal values in the fault data sample set and the normal data sample set by adopting an LOF algorithm;
S4, expanding minority fault sample data;
S5, extracting a plurality of types of normal sample data;
S6, training a model; integrating the normal sample and the fault sample after the secondary sampling to obtain a new balance sample set, inputting the new balance sample set into a random forest model for training, establishing a fault diagnosis model,
S7, classifying fault types of the transformer;
The preprocessing in S2 comprises normalization processing of the unbalanced data set, so that the numerical range of the data set is reduced, and the normalization method is as follows:
The initial dataset is brought into a normalized formula calculation, which is as follows:
Where z ij represents the original data, Representing the normalized data; i represents the number of sample data; j represents a characteristic dimension sequence number, and 1-5 represent five gas type data respectively;
the process of S3 includes:
s3.1, calculating sample point spacing: for n samples to be tested, the distance between any two sample points is calculated by adopting the mahalanobis distance, and the mahalanobis distance formula is as follows:
The reachable distance between the sample point O to be measured and the kth local cluster point is calculated, and the formula is as follows:
distk(O,P)＝MAX{distk(O),dist(O,P)}
the right side of the equation shows that the result takes the maximum value of the distance between the point and the point O and the distance between the kth cluster point P and the point O;
s3.2, calculating local reachable density lrd (O) of the sample point O to be detected:
Wherein N k (O) represents the kth distance field of the point O, namely the kth distance of the point O and all points within the kth distance, including the kth distance, |N k (O) | represents the number of the kth field points of the point O, and d k (O, P) represents the distance between the sample point O to be detected and the kth cluster point P;
S3.3, calculating local outlier factors LOF, wherein the LOF value reflects the difference between the reachable density of a certain point and the density of points in other fields and is used as a basis for judging whether the point O to be measured is an abnormal value or not:
In the formula, lrd (P) and lrd (O) respectively represent local reachable densities of the point P and the point O, if the LOF value is closer to 1, the reachable density of the point O is similar to the local reachable density of the point O, if the LOF value is smaller than 1, the reachable density of the point O is higher than the local reachable density of the point O, and if the LOF value is larger than 1, the reachable density of the point O is lower than the local reachable density of the point O, namely the reachable density is considered as an abnormal value;
The S4 is subjected to oversampling by adopting an SMOTE algorithm; the method comprises the following steps:
s4.1, setting the number of fault samples as T, calculating the sample spacing by using Euclidean distance, k neighbor samples x i(near) are obtained from the failure samples, near e1, once again, k }.
S4.2, setting a sampling multiplying power N according to an unbalanced proportion, randomly selecting one sample x i(nn) from k neighbor samples of each original fault sample x i, and combining the original fault sample and the neighbor fault samples into a new fault sample according to the following formula:
xnew＝xi+rand(0,1)*|xi-xi(nn)|
s4.3 repeating the two steps for N times to synthesize N new fault samples x i(new), new E { 1.. The N }, wherein the sample set at the moment is the new fault sample set;
the specific process of S5 comprises the following steps:
S5.1, firstly, clustering a data set into a plurality of clusters, and dividing a normal sample set into K clusters by using a K-means algorithm;
S5.2 randomly selecting M normal samples in each cluster group which has been clustered to form a subset X +, wherein
S5.3, calculating the information potential V (X) of each X +, wherein the calculation formula is shown as follows,Representing a gaussian kernel function:
s5.4, calculating the cross information potential V (X, Y) between each normal sample, wherein the calculation formula is as follows:
S5.5, constructing a cost function and solving the minimum value of the cost function, and finding a sample subset X + capable of minimizing the cost function, wherein the sample subset is M samples representing normal samples after undersampling, and the formula is as follows:
J(X+)＝H2(X+)+λDCS(X+,X+)
Wherein H 2(X+) is Renyi quadratic entropy for measuring normal redundancy, D CS(X+,X+) is CS divergence for processing distortion in undersampled data, and lambda E R + is used for controlling the distortion degree of the sampled data, the variables have the following relation:
DCS(X,Y)＝2H2(X,Y)-H2(X)-H2(Y)
the cost function is optimized according to the relation to obtain the following formula:
Wherein the parameter lambda epsilon R + is mainly used for controlling the distortion degree in the subsampled data;
S5.6 updates the samples in subset X +, the iterative formula is as follows:
Where X +k∈X+ (k e {1,2,., M }) refers to the samples in subset X +, Representing the gaussian kernel function,T epsilon N means that t times of iteration are performed;
s5.7, repeating the steps (3) to (6) until the result converges, so that M most representative normal samples are obtained;
The K-means algorithm in S5.1 comprises the following steps:
s5.1.1, firstly selecting K normal points as initial centroids;
S5.1.2 calculating the Euclidean distance from the point of the centroid neighbor to the centroid, and assigning each neighbor point to the nearest centroid to form K clusters;
s5.1.3 updating the mass center of each cluster according to the divided clusters; finally, repeating the first two steps until the mass center of the cluster does not move;
The fault types in S7 include: low energy discharge (D 1), high energy discharge (D 2), medium low temperature superheat (T 1), high temperature superheat (T 2) and normal data (NC).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210064482.1A CN114528921B (en) | 2022-01-20 | 2022-01-20 | Transformer fault diagnosis method based on LOF algorithm and mixed sampling |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210064482.1A CN114528921B (en) | 2022-01-20 | 2022-01-20 | Transformer fault diagnosis method based on LOF algorithm and mixed sampling |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114528921A CN114528921A (en) | 2022-05-24 |
| CN114528921B true CN114528921B (en) | 2024-06-11 |
Family
ID=81620371
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210064482.1A Active CN114528921B (en) | 2022-01-20 | 2022-01-20 | Transformer fault diagnosis method based on LOF algorithm and mixed sampling |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114528921B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116879662B (en) * | 2023-09-06 | 2023-12-08 | 山东华尚电气有限公司 | Transformer fault detection method based on data analysis |
| CN117349781B (en) * | 2023-12-06 | 2024-03-22 | 东莞市郡嘉电子科技有限公司 | Intelligent diagnosis method and system for faults of transformer |
| CN118011223B (en) * | 2024-04-08 | 2024-06-11 | 四川乐电新能源科技有限公司 | Fault diagnosis method and system for electric power energy storage system |
| CN118052540A (en) * | 2024-04-16 | 2024-05-17 | 北汽利戴工业技术服务(北京)有限公司 | Intelligent management method and system for equipment maintenance scheme |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107784325A (en) * | 2017-10-20 | 2018-03-09 | 河北工业大学 | Spiral fault diagnosis model based on the fusion of data-driven increment |
| CN110175640A (en) * | 2019-05-22 | 2019-08-27 | 镇江四联机电科技有限公司 | A kind of Fault Diagnosis Method of Electro-hydraulic based on machine learning |
| CN110398650A (en) * | 2019-08-05 | 2019-11-01 | 国网河北省电力有限公司电力科学研究院 | Based on k- adjacent to the Diagnosis Method of Transformer Faults of SMOTE and deep learning |
| CN113702728A (en) * | 2021-07-12 | 2021-11-26 | 广东工业大学 | Transformer fault diagnosis method and system based on combined sampling and LightGBM |
| CN113741364A (en) * | 2021-08-26 | 2021-12-03 | 上海应用技术大学 | Multi-mode chemical process fault detection method based on improved t-SNE |
| CN113866684A (en) * | 2021-11-14 | 2021-12-31 | 广东电网有限责任公司江门供电局 | Distribution transformer fault diagnosis method based on hybrid sampling and cost sensitivity |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107941537B (en) * | 2017-10-25 | 2019-08-27 | 南京航空航天大学 | A kind of mechanical equipment health state evaluation method |
-
2022
- 2022-01-20 CN CN202210064482.1A patent/CN114528921B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107784325A (en) * | 2017-10-20 | 2018-03-09 | 河北工业大学 | Spiral fault diagnosis model based on the fusion of data-driven increment |
| CN110175640A (en) * | 2019-05-22 | 2019-08-27 | 镇江四联机电科技有限公司 | A kind of Fault Diagnosis Method of Electro-hydraulic based on machine learning |
| CN110398650A (en) * | 2019-08-05 | 2019-11-01 | 国网河北省电力有限公司电力科学研究院 | Based on k- adjacent to the Diagnosis Method of Transformer Faults of SMOTE and deep learning |
| CN113702728A (en) * | 2021-07-12 | 2021-11-26 | 广东工业大学 | Transformer fault diagnosis method and system based on combined sampling and LightGBM |
| CN113741364A (en) * | 2021-08-26 | 2021-12-03 | 上海应用技术大学 | Multi-mode chemical process fault detection method based on improved t-SNE |
| CN113866684A (en) * | 2021-11-14 | 2021-12-31 | 广东电网有限责任公司江门供电局 | Distribution transformer fault diagnosis method based on hybrid sampling and cost sensitivity |
Non-Patent Citations (2)
| Title |
|---|
| 基于四阶段预处理与GBDT的油浸式变压器故障诊断方法;廖伟涵;郭创新;金宇;龚霄;;电网技术;20190128(第06期);全文 * |
| 基于失衡样本特性过采样算法与SVM的滚动轴承故障诊断;黄海松;魏建安;任竹鹏;吴江进;;振动与冲击;20200528(第10期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114528921A (en) | 2022-05-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114528921B (en) | Transformer fault diagnosis method based on LOF algorithm and mixed sampling | |
| CN112699913B (en) | Method and device for diagnosing abnormal relationship of household transformer in transformer area | |
| CN110579709B (en) | Fault diagnosis method for proton exchange membrane fuel cell for tramcar | |
| CN110994604B (en) | Power system transient stability assessment method based on LSTM-DNN model | |
| CN111478314B (en) | Transient stability evaluation method for power system | |
| CN113673129A (en) | Multi-objective robust optimization design method for ultra-high efficiency induction motor | |
| CN110609477B (en) | Electric power system transient stability discrimination system and method based on deep learning | |
| CN108875783A (en) | A kind of extreme learning machine Diagnosis Method of Transformer Faults towards unbalanced dataset | |
| CN113659565B (en) | Online prediction method for frequency situation of new energy power system | |
| CN109066651B (en) | Method for calculating limit transmission power of wind power-load scene | |
| CN110738232A (en) | grid voltage out-of-limit cause diagnosis method based on data mining technology | |
| CN112288147B (en) | Method for predicting insulation state of generator stator by BP-Adaboost strong predictor | |
| CN109188905B (en) | A kind of on-line monitoring method of the sound feature Cooperative Analysis towards gigawatt extra-supercritical unit | |
| CN115618249A (en) | Low-voltage power distribution station area phase identification method based on LargeVis dimension reduction and DBSCAN clustering | |
| Yin et al. | Deep learning based feature reduction for power system transient stability assessment | |
| CN111401792A (en) | Dynamic safety assessment method based on extreme gradient lifting decision tree | |
| CN109669414B (en) | Dynamic process monitoring method based on autocorrelation characteristic decomposition | |
| CN116680532A (en) | Transformer fault online diagnosis method for processing unbalanced small sample based on NNTR-SMOTE oversampling | |
| CN110348489B (en) | Transformer partial discharge mode identification method based on self-coding network | |
| CN116131261A (en) | Micro-grid dynamic equivalent modeling method considering model robustness | |
| CN114781551B (en) | Battery multi-fault intelligent classification and identification method based on big data | |
| Wu et al. | Research on Transformer Fault Diagnosis Based on SMOTE and Random Forest | |
| CN113341926B (en) | Multi-stage intermittent process fault detection method based on sparse weighted neighborhood preserving embedding | |
| Ren et al. | Application of RBF neural network optimized based on K-means cluster algorithm in fault diagnosis | |
| CN108508866B (en) | Million kilowatt unit fault variable identification method based on sparse relative discriminant analysis |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |