CN111239521B - Wind power generation system converter fault identification method based on PCA-kNN - Google Patents
Wind power generation system converter fault identification method based on PCA-kNN Download PDFInfo
- Publication number
- CN111239521B CN111239521B CN202010099166.9A CN202010099166A CN111239521B CN 111239521 B CN111239521 B CN 111239521B CN 202010099166 A CN202010099166 A CN 202010099166A CN 111239521 B CN111239521 B CN 111239521B
- Authority
- CN
- China
- Prior art keywords
- sample
- output voltage
- knn
- pca
- frequency domain
- 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
Images
Classifications
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Medical Informatics (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Control Of Eletrric Generators (AREA)
Abstract
The invention relates to a fault identification method for a converter of a wind power generation system based on PCA-kNN.
Description
Technical Field
The invention relates to the field of converter fault identification, in particular to a method for identifying converter faults based on a PCA-kNN wind power generation system.
Background
Wind power is an important energy source supplied by the world as a renewable energy source. With the continuous development of economy and science and technology, wind power generation is greatly developed in China, however, the working state of a fan cannot be stable due to the randomness of wind, so that all components of a wind power generation system frequently break down, a double Pulse Width Modulation (PWM) converter is used as an interface of the power generation system and a power grid, namely, the voltage is controlled to be stable, the maximum wind energy utilization rate is realized in a matching mode, and meanwhile, the converter is also one of the components which are most prone to break down.
The IGBT module is an important component of the converter, and due to the reasons of drive failure, switch damage and the like of the wind power generation system, the converter fault of the wind power generation system mainly considers the fault of an IGBT device, and the short-circuit fault of the IGBT can be monitored by the drive module integrated protection circuit at present, however, the open-circuit fault identification technology of the IGBT is still under study.
In 1968, Cover and Hart proposed a K-nearest neighbor (kNN) classification algorithm, which is one of the simplest classification algorithms in a data statistical model. The kNN algorithm idea is to calculate the distance between the measured sample and the known sample, and then determine the category of the measured sample according to the nearest k known samples. Because the kNN algorithm needs to calculate the distances between the detected samples and all the samples, when the characteristics extracted by the samples are more, the calculation amount of the kNN algorithm is more, and the calculation time is longer. Principal Component Analysis (PCA) is a common data dimension reduction and feature extraction method and is widely applied to the field of pattern recognition such as face recognition, so that the calculation amount of the algorithm can be reduced and the operation speed and recognition accuracy of the algorithm can be improved by combining the PCA algorithm with the kNN algorithm.
Disclosure of Invention
To solve the above existing problems. The invention provides a fault identification method for a converter of a wind power generation system based on PCA-kNN, which solves the problem of fault identification of the converter in the wind power generation system. To achieve this object:
the invention provides a PCA-kNN-based wind power generation system converter fault identification method, which comprises the following specific steps of:
step 1: detecting direct-current side output voltage signals of a back-to-back type three-phase PWM rectifier in the wind power generation system under various open-circuit faults;
step 2: extracting time domain, frequency domain and time-frequency domain characteristics of the output voltage, and establishing a sample set by the extracted characteristics;
and step 3: carrying out PCA (principal component analysis) processing on the characteristics of the sample set to obtain a characteristic vector of the output voltage signal after dimensionality reduction; arranging the reduced features according to the importance sequence from large to small, and selecting n-dimensional features retaining 99% of main information;
and 4, step 4: dividing a sample set into training samples and testing samples, taking n-dimensional features obtained after PCA calculation as input, calculating the distance from the testing samples to the training samples by using a kNN algorithm, and selecting the optimal k value in the kNN algorithm through cross validation;
and 5: detecting a direct current side output voltage signal of the rectifier to be detected, repeating the step 2 and the step 3, and finally classifying the rectifier faults by using a kNN classification algorithm.
As a further improvement of the present invention, the open-circuit faults of the rectifier in step 1 are as follows:
the open-circuit fault of the IGBT element of the PWM rectifier comprises the following steps: the single IGBT element open-circuit fault, 2 IGBT element open-circuit faults on the upper part and the lower part of the same bridge arm, 2 IGBT element open-circuit faults in the same half bridge and 2 IGBT element open-circuit faults on the upper part and the lower part of different bridge arms.
As a further improvement of the present invention, the time domain, frequency domain and time-frequency domain characteristics of the output voltage in step 2 are as follows:
the extracted time domain, frequency domain and time-frequency domain features include: area, energy, power spectrum estimation, wavelet coefficient sum of squares;
output voltage area: integration of voltage signal over time
Where u (n) is a voltage signal;
output voltage energy: sum of squares of voltage signal amplitudes
Output voltage power spectrum estimation:
where e (n) is white noise with a mean of 0 and a variance of σ2P is the order of the AR model, H (z) is the transfer function of the AR model system;
sum of squares of wavelet coefficients of output voltage: performing wavelet decomposition and reconstruction on the output voltage signal, wherein the decomposition layer number is 4, and db6 wavelet is selected
Wherein a is1、a2、a3、a4Is a low frequency approximate component of the signal, d1、d2、d3、d4Is the high frequency detail component of the signal.
As a further improvement of the present invention, in the step 3, PCA processing is performed on the characteristics of the sample set as follows:
for a given sample time, frequency and time-frequency domain feature set D ═ x1,x2,...,xmLet its projection coordinate in the low-dimensional coordinate system be Zi=(Zi1,Zi2,...,Zid) And has:
wherein ω isjIs a set of orthonormal vector bases, which can be based on Z according to equation 9iTo obtain xiOf the reconstructed sample
Original sample xiAnd reconstructed samples obtained based on projectionThe distance between can be expressed as:
where const is a constant, and W ═ ω1,ω2,...,ωd) According to the recent reconstructability principle, the minimization operation is performed on equation 10, since ω isjIs a basis of an orthonormal vector and,is a covariance matrix, the optimization objective of principal component analysis of equation 11 can be obtained:
wherein I represents an identity matrix, and formula 11 is transformed using lagrange multipliers to obtain formula 12:
XXTωi=λiωi (12)
final pair covariance matrix XXTDecomposing the eigenvalues, sorting the obtained eigenvalues in the order from small to large, and finally taking the first n eigenvectors corresponding to the maximum eigenvalue containing 99% of information to form a new eigenvector matrix Wopt=(ω1,ω2,…,ωn) Is the optimal solution of principal component analysis.
As a further improvement of the present invention, the distance from the test sample to the training sample is calculated by the kNN algorithm in step 4 as follows:
wherein xiIs a test characteristic sample, x'iAre training set feature samples.
As a further improvement of the present invention, in the step 4, the optimal k value in the kNN algorithm is selected as follows:
finding and testing sample x in training sample setiThe nearest k points, the field covering the k training sample points is denoted as Nk(x) Determining the class of the test sample according to the classification decision rule
Wherein, yi∈{c1,c2,...,ckIs a training sample x'iClass of instance, I is an indicator function when yi=ciIf so, I is 1, otherwise, I is 0; and finally, performing cross validation, and determining the k value according to the identification rate of the test sample.
The invention discloses a PCA-kNN wind power generation system-based converter fault identification method, which has the beneficial effects that:
1. the method extracts the characteristics of a time domain, a frequency domain and a time-frequency domain of a direct-current side output voltage signal, wherein the characteristics comprise more converter fault information;
2. the invention uses the PCA algorithm to reduce the dimension of the voltage signal characteristics, reduces the redundancy among the characteristics, improves the calculation speed of the kNN algorithm and reduces the classification time;
3. according to the method, the kNN algorithm is applied to converter fault identification, so that the accuracy and efficiency of converter fault identification are improved;
4. the invention provides an important technical means for identifying the fault of the converter of the wind power generation system.
Drawings
FIG. 1 is a flow chart of the overall algorithm principle;
FIG. 2 is a kNN classification schematic;
fig. 3 is a diagram of the converter fault identification result.
Detailed Description
The invention is described in further detail below with reference to the following detailed description and accompanying drawings:
the invention provides a PCA-kNN-based wind power generation system converter fault identification method, which extracts the characteristics of time domain, frequency domain and time-frequency domain of direct-current voltage at the direct-current side of a wind power generation system converter, performs dimension reduction processing on the characteristics by using the dimension reduction capability of the PCA, and finally classifies samples to be detected by using a kNN algorithm. The overall algorithm principle flow of the invention is shown in fig. 1, and the specific steps are as follows:
step 1: detecting direct-current side output voltage signals of a back-to-back type three-phase PWM rectifier in the wind power generation system under various open-circuit faults;
the open-circuit faults of the rectifier in the step 1 are specifically described as follows:
the open-circuit fault of the IGBT element of the PWM rectifier comprises the following steps: the single IGBT element open-circuit fault, 2 IGBT element open-circuit faults on the upper part and the lower part of the same bridge arm, 2 IGBT element open-circuit faults in the same half bridge and 2 IGBT element open-circuit faults on the upper part and the lower part of different bridge arms.
Step 2: extracting time domain, frequency domain and time-frequency domain characteristics of the output voltage, and establishing a sample set by the extracted characteristics;
the time domain, frequency domain and time-frequency domain characteristics of the output voltage in step 2 are described as follows:
the extracted time domain, frequency domain and time-frequency domain features include: area, energy, power spectrum estimation, wavelet coefficient sum of squares;
output voltage area: integration of voltage signal over time
Wherein u (n) is a voltage signal;
output voltage energy: sum of squares of voltage signal amplitudes
Output voltage power spectrum estimation:
where e (n) is white noise with a mean of 0 and a variance of σ2P is the order of the AR model, H (z) is the transfer function of the AR model system;
sum of squares of wavelet coefficients of output voltage: performing wavelet decomposition and reconstruction on the output voltage signal, wherein the decomposition layer number is 4, and db6 wavelet is selected
Wherein a is1、a2、a3、a4Is a low frequency approximate component of the signal, d1、d2、d3、d4Is the high frequency detail component of the signal.
And step 3: carrying out PCA (principal component analysis) processing on the characteristics of the sample set to obtain a characteristic vector of the output voltage signal after dimensionality reduction; arranging the reduced features according to the importance sequence from large to small, and selecting n-dimensional features retaining 99% of main information;
the PCA treatment of the features of the sample set in step 3 is described in detail as follows:
for a given sample time, frequency and time-frequency domain feature set D ═ x1,x2,...,xmLet its projection coordinate in the low-dimensional coordinate system be Zi=(Zi1,Zi2,...,Zid) And has:
wherein ω isjIs a set of orthonormal vector bases, which can be based on Z according to equation 9iTo obtain xiOf the reconstructed sample
Original sample xiAnd reconstructed samples obtained based on projectionThe distance between can be expressed as:
where const is a constant, and W ═ ω1,ω2,…,ωd) According to the recent reconstructability principle, the minimization operation is performed on equation 10, since ω isjIs a basis of an orthonormal vector and,is a covariance matrix, the optimization objective of principal component analysis of equation 11 can be obtained:
wherein I represents an identity matrix, and formula 11 is transformed using lagrange multipliers to obtain formula 12:
XXTωi=λiωi (12)
final pair covariance matrix XXTDecomposing the eigenvalues, sorting the obtained eigenvalues in the order from small to large, and finally taking the first n eigenvectors corresponding to the maximum eigenvalues containing 99% of information to form a new eigenvector matrix Wopt=(ω1,ω2,…,ωn) Is the optimal solution of principal component analysis.
And 4, step 4: dividing a sample set into training samples and testing samples, taking n-dimensional features obtained after PCA calculation as input, calculating the distance from the testing samples to the training samples by using a kNN algorithm, and selecting the optimal k value in the kNN algorithm through cross validation;
the calculation of the distance from the test sample to the training sample by the kNN algorithm in step 4 is specifically described as follows:
wherein xiIs a test characteristic sample, x'iAre training set feature samples.
The selection of the optimal k value in the kNN algorithm in step 4 is specifically described as follows:
finding and testing sample x in training sample setiThe nearest k points, the field covering the k training sample points is denoted as Nk(x) As shown in FIG. 2, the classification of the test sample is determined according to the classification decision rule
Wherein, yi∈{c1,c2,…,ckIs a training sample x'iClass of instance, I is an indicator function when yi=ciIf so, I is 1, otherwise, I is 0; and finally, performing cross validation, and determining the k value according to the identification rate of the test sample.
And 5: detecting a direct current side output voltage signal of the rectifier to be detected, repeating the step 2 and the step 3, and finally classifying the rectifier faults by using a kNN classification algorithm, wherein the classification result is shown in fig. 3.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention in any way, but any modifications or equivalent variations made according to the technical spirit of the present invention are within the scope of the present invention as claimed.
Claims (2)
1. The method for identifying the converter fault of the wind power generation system based on PCA-kNN comprises the following steps,
step 1: detecting direct-current side output voltage signals of a back-to-back type three-phase PWM rectifier in the wind power generation system under various open-circuit faults;
the various open-circuit faults of the rectifier in the step 1 are as follows:
the open-circuit fault of the IGBT element of the PWM rectifier comprises the following steps: the single IGBT element open-circuit fault, 2 IGBT element open-circuit faults on the upper part and the lower part of the same bridge arm, 2 IGBT element open-circuit faults in the same half bridge, and 2 IGBT element open-circuit faults on the upper part and the lower part of different bridge arms;
step 2: extracting time domain, frequency domain and time-frequency domain characteristics of the output voltage, and establishing a sample set by the extracted characteristics;
the time domain, frequency domain and time-frequency domain characteristics of the output voltage in step 2 are as follows:
the extracted time domain, frequency domain and time-frequency domain features include: area, energy, power spectrum estimation, wavelet coefficient sum of squares;
output voltage area: integration of voltage signal over time
Where u (n) is a voltage signal;
output voltage energy: sum of squares of voltage signal amplitudes
Output voltage power spectrum estimation:
where e (n) is white noise with a mean of 0 and a variance of σ2P is the order of the AR model, H (z) is the transfer function of the AR model system;
sum of squares of wavelet coefficients of output voltage: performing wavelet decomposition and reconstruction on the output voltage signal, wherein the decomposition layer number is 4, and db6 wavelet is selected
Wherein a is1、a2、a3、a4Is a low frequency approximate component of the signal, d1、d2、d3、d4Is the high frequency detail component of the signal;
and step 3: carrying out PCA (principal component analysis) processing on the characteristics of the sample set to obtain a characteristic vector of the output voltage signal after dimensionality reduction; arranging the reduced features according to the importance sequence from large to small, and selecting n-dimensional features retaining 99% of main information;
the PCA treatment of the sample set features in step 3 is as follows:
for a given sample time, frequency and time-frequency domain feature set D ═ x1,x2,…,xmLet its projection coordinate in the low-dimensional coordinate system be Zi=(Zi1,Zi2,...,Zid) And has:
wherein ω isjIs a set of orthonormal vector bases, which can be based on Z according to equation 9iTo obtain xiOf the reconstructed sample
Original sample xiAnd reconstructed samples obtained based on projectionThe distance between can be expressed as:
where const is a constant, and W ═ ω1,ω2,...,ωd) According to the recent reconstructability principle, the minimization operation is performed on equation 10, since ω isjIs a basis of an orthonormal vector and,is a covariance matrix, the optimization objective of principal component analysis of equation 11 can be obtained:
wherein I represents an identity matrix, and formula 11 is transformed using lagrange multipliers to obtain formula 12:
XXTωi=λiωi (12)
final pair covariance matrix XXTDecomposing the eigenvalues, sorting the obtained eigenvalues in the order from small to large, and finally taking the first n eigenvectors corresponding to the maximum eigenvalue containing 99% of information to form a new eigenvector matrix Wopt=(ω1,ω2,...,ωn) The optimal solution of the principal component analysis is obtained;
and 4, step 4: dividing a sample set into training samples and testing samples, taking n-dimensional features obtained after PCA calculation as input, calculating the distance from the testing samples to the training samples by using a kNN algorithm, and selecting the optimal k value in the kNN algorithm through cross validation;
in step 4, the kNN algorithm calculates the distances from the test samples to the training samples as follows:
wherein xiIs a test characteristic sample, x'iAre training set feature samples.
2. The PCA-kNN-based wind power generation system converter fault identification method of claim 1, wherein; the optimal k value in the kNN algorithm selected in step 4 is as follows:
finding and testing sample x in training sample setiThe nearest k points, the field covering the k training sample points is denoted as Nk(x) Determining the class of the test sample according to the classification decision rule
Wherein, yi∈{c1,c2,...,ckIs a training sample x'iClass of instance, I is an indicator function when yi=ciIf so, I is 1, otherwise, I is 0; finally, cross validation is carried out, and a k value is determined according to the identification rate of the test sample;
and 5: detecting a direct current side output voltage signal of the rectifier to be detected, repeating the step 2 and the step 3, and finally classifying the rectifier faults by using a kNN classification algorithm.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010099166.9A CN111239521B (en) | 2020-02-18 | 2020-02-18 | Wind power generation system converter fault identification method based on PCA-kNN |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010099166.9A CN111239521B (en) | 2020-02-18 | 2020-02-18 | Wind power generation system converter fault identification method based on PCA-kNN |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111239521A CN111239521A (en) | 2020-06-05 |
CN111239521B true CN111239521B (en) | 2022-02-11 |
Family
ID=70879970
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010099166.9A Active CN111239521B (en) | 2020-02-18 | 2020-02-18 | Wind power generation system converter fault identification method based on PCA-kNN |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111239521B (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103018601A (en) * | 2012-12-05 | 2013-04-03 | 江南大学 | Primary fault diagnosis method of converter in wind turbine system |
CN105136454A (en) * | 2015-10-15 | 2015-12-09 | 上海电机学院 | Wind turbine gear box fault recognition method |
CN106646096A (en) * | 2016-11-15 | 2017-05-10 | 国网四川省电力公司广安供电公司 | Transformer fault classification and identification method based on vibration analysis method |
CN108073158A (en) * | 2017-12-05 | 2018-05-25 | 上海电机学院 | Based on PCA and KNN density algorithm Wind turbines Method for Bearing Fault Diagnosis |
CN109164343A (en) * | 2018-08-30 | 2019-01-08 | 西华大学 | Quantify the Diagnosis Method of Transformer Faults with weighting KNN based on characteristic information |
-
2020
- 2020-02-18 CN CN202010099166.9A patent/CN111239521B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103018601A (en) * | 2012-12-05 | 2013-04-03 | 江南大学 | Primary fault diagnosis method of converter in wind turbine system |
CN105136454A (en) * | 2015-10-15 | 2015-12-09 | 上海电机学院 | Wind turbine gear box fault recognition method |
CN106646096A (en) * | 2016-11-15 | 2017-05-10 | 国网四川省电力公司广安供电公司 | Transformer fault classification and identification method based on vibration analysis method |
CN108073158A (en) * | 2017-12-05 | 2018-05-25 | 上海电机学院 | Based on PCA and KNN density algorithm Wind turbines Method for Bearing Fault Diagnosis |
CN109164343A (en) * | 2018-08-30 | 2019-01-08 | 西华大学 | Quantify the Diagnosis Method of Transformer Faults with weighting KNN based on characteristic information |
Non-Patent Citations (1)
Title |
---|
基于PCA-KNN聚类的通用在线故障诊断算法设计;赵晓君等;《计算机测量与控制》;20151231;第23卷(第8期);第2762-2765页 * |
Also Published As
Publication number | Publication date |
---|---|
CN111239521A (en) | 2020-06-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Lin et al. | Detection and classification of multiple power-quality disturbances with wavelet multiclass SVM | |
Cui et al. | A novel approach of analog circuit fault diagnosis using support vector machines classifier | |
Strangas et al. | Time–frequency analysis for efficient fault diagnosis and failure prognosis for interior permanent-magnet AC motors | |
Manimala et al. | Optimization techniques for improving power quality data mining using wavelet packet based support vector machine | |
Ondel et al. | A method to detect broken bars in induction machine using pattern recognition techniques | |
CN110068776B (en) | Three-level inverter open-circuit fault diagnosis method based on optimized support vector machine | |
CN106548013B (en) | Utilize the voltage sag source identification method for improving incomplete S-transformation | |
CN109359665B (en) | Household appliance load identification method and device based on support vector machine | |
CN111080105A (en) | Transformer area user-to-user relationship identification method and system based on voltage time sequence data | |
Gargoom et al. | Investigation of effective automatic recognition systems of power-quality events | |
CN109858564B (en) | Improved Adaboost-SVM model generation method suitable for wind power converter fault diagnosis | |
CN114359674A (en) | Non-invasive load identification method based on metric learning | |
CN111965486A (en) | Power distribution network fault identification method and system based on intelligent data fusion analysis | |
CN105572501A (en) | Power quality disturbance identification method based on SST conversion and LS-SVM | |
CN113125992B (en) | NPC three-level inverter fault diagnosis method and system based on DBN | |
CN110705694A (en) | Electric larceny monitoring method facing edge data center based on feature extraction | |
CN111856209A (en) | Power transmission line fault classification method and device | |
CN114881077A (en) | Voltage sag source classification method and system based on time sequence trajectory characteristics | |
CN113702767B (en) | Island direct-current microgrid fault diagnosis method based on wavelet sliding window energy | |
CN115688880A (en) | Open-circuit fault diagnosis method and device for switching device of modular multilevel converter | |
Zhang et al. | Towards multi-scenario power system stability analysis: An unsupervised transfer learning method combining DGAT and data augmentation | |
Kumar et al. | Accurate fault diagnosis and classification scheme based on non-parametric, statistical-frequency features and neural networks | |
CN111239521B (en) | Wind power generation system converter fault identification method based on PCA-kNN | |
CN112636328B (en) | Medium-voltage distribution network equipment load state identification method | |
Cui et al. | T-type inverter fault diagnosis based on GASF and improved AlexNet |
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 | ||
TR01 | Transfer of patent right | ||
TR01 | Transfer of patent right |
Effective date of registration: 20230630 Address after: Room 1028, Nanyuan Building, No. 58, Nanhu Road, Jianye District, Nanjing, Jiangsu Province, 210017 Patentee after: Nanjing Waliang Technology Co.,Ltd. Address before: No. 99 Jiangning Road, Nanjing District hirokage 210000 cities in Jiangsu Province Patentee before: JINLING INSTITUTE OF TECHNOLOGY |