CN110705581A - Pitch bearing fault identification method based on improved hidden Markov model - Google Patents
Pitch bearing fault identification method based on improved hidden Markov modelDownload PDFInfo
- Publication number
- CN110705581A CN110705581ACN201910650675.3ACN201910650675ACN110705581ACN 110705581 ACN110705581 ACN 110705581ACN 201910650675 ACN201910650675 ACN 201910650675ACN 110705581 ACN110705581 ACN 110705581A
- Authority
- CN
- China
- Prior art keywords
- pitch bearing
- hidden markov
- markov model
- fault
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/29—Graphical models, e.g. Bayesian networks
- G06F18/295—Markov models or related models, e.g. semi-Markov models; Markov random fields; Networks embedding Markov models
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
- G01M13/045—Acoustic or vibration analysis
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Acoustics & Sound (AREA)
- Wind Motors (AREA)
Abstract
Description
Technical Field
The invention belongs to the field of wind power generation, and particularly relates to a variable pitch bearing fault identification method based on an improved hidden Markov model.
Background
The variable-pitch bearing is used as one of core components of a control system of a modern large variable-speed constant-frequency wind generating set, and plays an important role in safe and stable operation of the set. However, due to randomness and uncertainty of wind speed, and the pitch bearing works in the hub along with rotation of the blades, the severe operating environment makes the pitch bearing one of the components with high failure rate in the wind turbine, and once the pitch bearing fails or is abnormal, serious accidents of the wind turbine such as blade fracture and fan collapse are likely to occur. In addition, the on-line monitoring parameters of the variable pitch bearing are numerous, and the operation parameters are closely related to the complex operation conditions of the wind turbine generator, so that the accurate on-line monitoring of the state of the variable pitch bearing is very difficult. Therefore, the abnormal operation state of the variable pitch bearing is reasonably and accurately identified, and the method has important practical significance for effectively avoiding the occurrence of faults and cascading faults of the variable pitch bearing, improving the state evaluation level of the wind turbine generator and realizing the efficient and reliable grid-connected operation of the wind turbine generator.
The randomness of the change of the power and the rotating speed of the wind turbine generator is large, and the variable-pitch bearing rotates at a low speed or does not rotate completely, so that the effect of a vibration analysis method for identifying the state of the variable-pitch bearing of the wind turbine generator is limited. Because the variable-pitch bearing is mostly lubricated by lubricating grease or mixed lubrication of the lubricating grease and lubricating oil, an online monitoring method is difficult to adopt; during off-line analysis, it is also difficult to ensure that the collected sample participates in the lubricating work, so that the effect of the oil monitoring method for identifying the state of the variable-pitch bearing of the fan is also difficult to ensure. Therefore, in recent years, a pitch bearing fault monitoring method based on SCADA data of a monitoring system is slowly developed.
However, in the existing fault monitoring method based on the variable pitch bearing, all faults in the operation process of the wind turbine generator are assumed to be known in the modeling stage, but with the rapid development of the wind power generation technology, the wind turbine generator is increasingly large and complicated, faults which do not appear before may occur in the operation process, and if the faults are not considered, the fault accurate identification of the variable pitch bearing is very unfavorable.
Disclosure of Invention
The invention aims to provide a wind generating set variable pitch bearing fault identification method based on an improved hidden Markov model aiming at the problem that the online monitoring of the state of a variable pitch bearing is difficult, and the method considers that besides known faults, unknown faults which are not considered in the running process of a wind generating set are also very likely to occur and need extra attention. Thus, a threshold statistic is defined based on a hidden Markov model for the identification of unknown faults. According to the invention, a new threshold value statistic is introduced, so that the false alarm rate can be greatly reduced when the fault of the variable-pitch bearing of the wind generating set is identified on line. The method has higher accuracy for the fault identification of the variable-pitch bearing of the wind generating set.
In order to achieve the purpose, the invention adopts the technical scheme that: a wind generating set variable pitch bearing fault identification method based on an improved hidden Markov model comprises the following steps:
step 1: off-line modeling, wherein training sample set data is collected, and a hidden Markov model is trained by using the collected training set data;
step 2: calculating a threshold value for identifying unknown faults by using the trained hidden Markov model;
and step 3: performing online identification, namely taking the online collected test data as a monitoring sample, and calculating the posterior probability variance of the test data;
and 4, step 4: and comparing the posterior probability variance of the calculated test data with a threshold value, and judging the fault type of the variable pitch bearing.
The step 1 offline modeling process comprises the following steps:
step 101: suppose that the monitoring data collected in the running process of the wind turbine set form X ═ X1，x2，L，xn]T∈Rn×mWhere m denotes the number of monitored variables, n denotes the number of samples, xi∈RmI is 1, i represents the ith sample, i is 1, L, n;
step 102: training the hidden Markov model by taking the training set data as an observation sequence O of the hidden Markov model to obtain a model parameter lambda (A, B, pi, M, N), wherein:
a is a hidden state transition probability matrix which describes the transition probability among all states in a hidden Markov model; b is an observed value probability density matrix; π is the initial state probability matrix; m is the number of gaussian mixture components per implicit state. N is the number of implicit states, which is the number of all the conditions in the training sample set.
A is a hidden state transition probability matrix describing the transition probability between states in a hidden Markov model, as shown in equations (1) to (2)
In the formula
Abbreviated as ai,jIndicating that at time t the hidden state is SiUnder the condition of (1), the hidden state is S at the moment of t +1jProbability of (a), qtRepresenting an implicit state at time t;
b is an observed value probability density matrix, and the specific calculation is shown in the following formula
Wherein O is an observation sequence, CjmIs a hidden state SjH is the log-concave or elliptical symmetric density, mujmIs a hidden state SjThe mean vector of the m-th mixture component, UjmIs a hidden state SjThe covariance matrix of the mth mixture component, CjmSatisfying the random constraint:
Cjm、μjm、Ujmthe reevaluation of (d) is such that: cjmThe reestimated value of (A) is a hidden Markov model in a hidden state SjExpectation of degree with k-th mixed component divided by hidden Markov model in hidden state SjExpectation of the number of times of (d), mujmIs the partial expectation, U, of the observation vector described by the k-th mixture componentjmIs the partial covariance of the observation vector described by the k-th mixture component, see equations (5) to (7),
wherein alpha ist(j)＝P(O1,O2,…,Ot,qt＝Sj| λ) indicates that the hidden state is S at time t, given the model parameter λ ═ a, B, pi, M, NjAnd the observation sequence is O1,O2,…,OtThe probability of (d); beta is at(j)＝P(Ot+1,Ot+2,…,OT|qt＝Sjλ) denotes the hidden state q at a given model parameter λ ═ (a, B, pi, M, N) and time tt＝SjUnder the condition (1), the sequence observed from T +1 to T is Ot+1,Ot+2,…,OTThe probability of (d);
is CjkIs determined by the estimated value of (c),
is mujkIs determined by the estimated value of (c),is UjkAn estimate of (d).
The process of calculating the threshold value for identifying the unknown fault by using the trained hidden Markov model comprises,
step 201: calculating the variance of the posterior probabilities of all training samplesVariance of posterior probability of each observation sample in training set
The following can be calculated:
step 202: taking the variance of the minimum a posteriori probability
As a threshold value
The threshold expression is:
preferably, the threshold value
The tolerance coefficient alpha is contained, and the value range of the tolerance coefficient alpha is between 0 and 1. The threshold expression is:
preferably, the data acquired on line in the operation process of the wind generating set comprises normal operation condition data of the wind generating set, fault condition data of a variable pitch bearing and another fault condition data of a variable pitch system of the wind generating set during operation.
Preferably, the off-line modeling and the acquisition of the training sample set data comprise normal operation condition data of the wind generating set and fault condition data of the pitch bearing.
Preferably, if the most recent observation is from a known condition, the Viterbi algorithm is used to determine the condition for the current sample.
Preferably, the wind turbine generator set has dynamic property in the power generation process.
The invention has the beneficial effects that: the invention provides a wind generating set variable-pitch bearing fault identification method based on an improved hidden Markov model for the first time, so that fault identification of a variable-pitch bearing is realized; the method can solve the problem of the dynamic property of the process data, so that the fault identification of the variable-pitch bearing can be more effectively carried out; the invention defines a threshold value statistic based on the hidden Markov model for identifying unknown faults, thereby greatly reducing the false alarm rate of fault identification of the variable pitch bearing.
Drawings
FIG. 1 is a flow chart of an embodiment of the present invention
Detailed Description
Example (b):
with reference to the flowchart in fig. 1, the method for identifying the fault of the pitch bearing of the wind generating set based on the improved hidden markov model provided by the invention comprises the following steps:
step 1: off-line modeling, collecting training sample set data including normal operation condition data of the wind generating set and fault condition data of a variable pitch bearing, and training a hidden Markov model by using the collected training set data;
step 2: calculating a threshold value for identifying unknown faults by using the trained hidden Markov model;
and step 3: and performing online identification, wherein the online collected test data comprises normal operation condition data of the wind generating set, fault condition data of a variable pitch bearing and another fault condition data of a variable pitch system of the wind generating set during operation. As a monitoring sample, calculating the posterior probability variance of the test data;
and 4, step 4: and comparing the posterior probability variance of the calculated test data with a threshold value, and judging the fault type of the variable pitch bearing.
The off-line modeling process described in step 1 is as follows:
step 101: suppose that the monitoring data collected in the running process of the wind turbine set form X ═ X1,x2,L,xn]T？Rn′mWhere m denotes the number of monitored variables, n denotes the number of samples, xi∈RmI represents the ith sample, i is 1, L, n;
step 102: training a hidden Markov model by taking training set data as an observation sequence O of the hidden Markov model to obtain a model parameter lambda (A, B, pi, M, N);
wherein, A is a hidden state transition probability matrix, B is an observed value probability density matrix, pi is an initial state probability matrix, M is the number of Gaussian mixture components in each hidden state, N is the number of hidden states, and the number of the hidden states is the number of all working conditions in a training sample set.
A is a hidden state transition probability matrix describing the transition probability between states in a hidden Markov model, as shown in equations (1) to (2)
In the formula
Abbreviated as ai,jIndicating that at time t the hidden state is SiUnder the condition of (1), the hidden state is S at the moment of t +1jProbability of (a), qtRepresenting an implicit state at time t;
b is an observed value probability density matrix, and the specific calculation is shown in the following formula
Wherein O is an observation sequence, CjmIs a hidden state SjH is the log-concave or elliptical symmetric density, mujmIs a hidden state SjThe mean vector of the m-th mixture component, UjmIs a hidden state SjThe covariance matrix of the mth mixture component, CjmSatisfying the random constraint:
Cjm、μjm、Ujmthe reevaluation of (d) is such that: cjmThe reestimated value of (A) is a hidden Markov model in a hidden state SjExpectation of degree with k-th mixed component divided by hidden Markov model in hidden state SjExpectation of the number of times of (d), mujmIs the partial expectation, U, of the observation vector described by the k-th mixture componentjmIs the partial covariance of the observation vector described by the k-th mixture component, see equations (5) to (7),
wherein alpha ist(j)＝P(O1,O2,…,Ot,qt＝Sj| λ) indicates that the hidden state is S at time t, given the model parameter λ ═ a, B, pi, M, NjAnd the observation sequence is O1,O2,…,OtThe probability of (d); beta is at(j)＝P(Ot+1,Ot+2,…,OT|qt＝Sjλ) denotes the hidden state q at a given model parameter λ ═ (a, B, pi, M, N) and time tt＝SjUnder the condition (1), the sequence observed from T +1 to T is Ot+1,Ot+2,…,OTThe probability of (d);
is CjkIs determined by the estimated value of (c),
is mujkIs determined by the estimated value of (c),is UjkAn estimate of (d).
Calculating a threshold value for identifying the unknown fault by using the trained hidden Markov model: in the process of calculating the hidden markov model parameter λ ═ (a, B, pi, M, N), an observation probability density distribution matrix can be obtained. Element b in the matrixj(Ot)＝P(Ot|qt＝Sj) Indicating that the hidden state at the moment t is SjTime observation sample OtThe probability of occurrence. For the observation sample, it is noted that the probability that it belongs to its own condition is much greater than the probability that it belongs to other conditions.
Step 2, the online identification process is as follows:
step 201: calculating the variance of the posterior probabilities of all training samples
Variance of posterior probability of each observation sample in training set
The following can be calculated:
step 202: taking the variance of the minimum a posteriori probability
As a threshold value
The threshold expression is:
the variance of the posterior probability is a relatively large value as long as the current observed sample belongs to one of the known operating conditions. Conversely, if the observation sample does not belong to any known operating condition, the probability that it belongs to each known operating condition will be very small. Therefore, its variance is an extremely small value. Based on the above analysis, the variance of the posterior probability of each observation sample can be used as an index to identify unknown faults. Since all the conditions in the training set are known, we can obtain the predetermined threshold P by minimizing the variance of the posterior probabilities of all the training samplest*. For samples in the test set data, if the variance of their posterior probability is less than Pt*It indicates that an unknown pattern is present in the process. The threshold value may be set according to the following equation:
wherein α is Pt*In fact in order to guarantee the threshold value Pt*The effectiveness of the (c),the tolerance factor α ranges from 0 to 1, which is a particularly small value.
The posterior probability variance is compared with a threshold value Pt*Comparing, if the variance of the posterior probability of the sample in the test set data is less than the threshold value Pt*If the variance of the posterior probability is larger than the threshold value P, the unknown fault is shown to appear in the processt*It indicates that the position fault is a known condition in the process.
If the latest observed value is from the known working condition, the Viterbi algorithm is used to determine the working condition corresponding to the current sample.
And if the latest observed value is from the unknown working condition, outputting the unknown fault, carrying out subsequent analysis on the unknown fault, analyzing the source of the unknown fault, solving the unknown fault, and putting the unknown fault serving as the known fault into an offline modeling step.
The wind turbine generator set has dynamic property in the power generation process, and can generate unknown faults.
The invention provides a wind generating set variable-pitch bearing fault identification method based on an improved hidden Markov model for the first time, so that fault identification of a variable-pitch bearing is realized; the method can solve the problem of the dynamic property of the process data, so that the fault identification of the variable-pitch bearing can be more effectively carried out; the invention defines a threshold value statistic based on the hidden Markov model for identifying unknown faults, thereby greatly reducing the false alarm rate of fault identification of the variable pitch bearing.
Claims (10)
1. A variable pitch bearing fault identification method based on an improved hidden Markov model is characterized by comprising the following steps:
step 1: off-line modeling, wherein training sample set data is collected, and a hidden Markov model is trained by using the collected training set data;
step 2: calculating a threshold value for identifying unknown faults by using the trained hidden Markov model;
and step 3: performing online identification, namely taking the online collected test data as a monitoring sample, and calculating the posterior probability variance of the test data;
and 4, step 4: and comparing the posterior probability variance of the calculated test data with a threshold value, and judging the fault type of the variable pitch bearing.
2. The method for identifying the fault of the variable pitch bearing based on the improved hidden Markov model as claimed in claim 1, wherein the off-line modeling process is as follows:
step 101: the method includes the steps that X is formed by training sample set data collected in the running process of a wind turbine generator (X ═ X)1，x2，L，xn]T∈Rn×mWhere m denotes the number of monitored variables, n denotes the number of samples, xi∈RmI represents the ith sample, i is 1, L, n;
step 102: training the hidden Markov model by taking the training set data as an observation sequence O of the hidden Markov model to obtain a model parameter lambda (A, B, pi, M, N), wherein:
a is a hidden state transition probability matrix which describes the transition probability among all states in a hidden Markov model;
b is an observed value probability density matrix;
π is the initial state probability matrix;
m is the number of Gaussian mixture components in each hidden state;
n is the number of implicit states, which is the number of all the conditions in the training sample set.
3. The pitch bearing fault identification method based on the improved hidden Markov model is characterized in that the threshold value process for identifying the unknown fault by utilizing the trained hidden Markov model comprises the following steps,
step 201: calculating the variance of the posterior probabilities of all training samples
Step 202: taking the variance of the minimum a posteriori probability
As a threshold value
The threshold expression is:
4. the method for identifying the fault of the variable-pitch bearing based on the improved hidden Markov model as claimed in claim 2, wherein the threshold value is setThe tolerance coefficient alpha is contained, and the value range of the tolerance coefficient alpha is between 0 and 1. The threshold expression is:
5. the improved hidden Markov model-based pitch bearing fault identification method according to claim 1, wherein the online collection of data of the wind turbine generator system during operation comprises normal operation condition data of the wind turbine generator system, fault condition data of the pitch bearing, and data of another fault condition occurring in a pitch system of the wind turbine generator system during operation.
6. The method for identifying the fault of the variable pitch bearing based on the improved hidden Markov model is characterized in that the offline modeling and the acquisition of the training sample set data comprise normal operation condition data of the wind generating set and fault condition data of the variable pitch bearing.
7. The method for identifying the fault of the variable pitch bearing based on the improved hidden Markov model as claimed in claim 1, wherein the posterior probability variance of the variable pitch bearing is compared with a threshold Pt*Comparing, if the variance of the posterior probability of the sample in the test set data is less than the threshold value Pt*If the variance of the posterior probability is larger than the threshold value P, the unknown fault is shown to appear in the processt*It indicates that the position fault is a known condition in the process.
8. The improved hidden Markov model-based pitch bearing fault identification method of claim 4, wherein if the latest observed value is from a known condition, a Viterbi algorithm is used to determine the condition corresponding to the current sample.
9. The method for identifying a fault in a pitch bearing based on an improved hidden markov model according to claim 1, wherein the wind turbine generator is dynamic.
10. Method for pitch bearing fault identification based on improved hidden markov models, according to claim 1, wherein the variance of the posterior probability
The following can be calculated:
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910650675.3ACN110705581A (en) | 2019-07-18 | 2019-07-18 | Pitch bearing fault identification method based on improved hidden Markov model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910650675.3ACN110705581A (en) | 2019-07-18 | 2019-07-18 | Pitch bearing fault identification method based on improved hidden Markov model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110705581Atrue CN110705581A (en) | 2020-01-17 |
Family
ID=69193164
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910650675.3APendingCN110705581A (en) | 2019-07-18 | 2019-07-18 | Pitch bearing fault identification method based on improved hidden Markov model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110705581A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113494428A (en)* | 2020-03-20 | 2021-10-12 | 新疆金风科技股份有限公司 | Fault detection method and device of wind generating set |
| CN113723546A (en)* | 2021-09-03 | 2021-11-30 | 江苏理工学院 | Bearing fault detection method and system based on discrete hidden Markov model |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104793606A (en)* | 2015-04-15 | 2015-07-22 | 浙江大学 | Industrial fault diagnosis method based on improved KPCA (kernel principal component analysis) and hidden Markov model |
| CN104897784A (en)* | 2015-03-15 | 2015-09-09 | 国家电网公司 | Transformer fault diagnosis method based on coupled hidden Markov model |
| CN106405384A (en)* | 2016-08-26 | 2017-02-15 | 中国电子科技集团公司第十研究所 | Simulation circuit health state evaluation method |
| CN107579846A (en)* | 2017-08-23 | 2018-01-12 | 昆明理工大学 | A cloud computing fault data detection method and system |
| CN108960328A (en)* | 2018-07-04 | 2018-12-07 | 温州大学 | A method of the diagnosis automobile hub bearing failure based on Hankel matrix |
| CN109597307A (en)* | 2018-12-29 | 2019-04-09 | 浙江运达风电股份有限公司 | A kind of failure tolerant control method based on Hidden Markov Model |
- 2019
- 2019-07-18CNCN201910650675.3Apatent/CN110705581A/enactivePending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104897784A (en)* | 2015-03-15 | 2015-09-09 | 国家电网公司 | Transformer fault diagnosis method based on coupled hidden Markov model |
| CN104793606A (en)* | 2015-04-15 | 2015-07-22 | 浙江大学 | Industrial fault diagnosis method based on improved KPCA (kernel principal component analysis) and hidden Markov model |
| CN106405384A (en)* | 2016-08-26 | 2017-02-15 | 中国电子科技集团公司第十研究所 | Simulation circuit health state evaluation method |
| CN107579846A (en)* | 2017-08-23 | 2018-01-12 | 昆明理工大学 | A cloud computing fault data detection method and system |
| CN108960328A (en)* | 2018-07-04 | 2018-12-07 | 温州大学 | A method of the diagnosis automobile hub bearing failure based on Hankel matrix |
| CN109597307A (en)* | 2018-12-29 | 2019-04-09 | 浙江运达风电股份有限公司 | A kind of failure tolerant control method based on Hidden Markov Model |
Non-Patent Citations (1)
| Title |
|---|
| 王琳: "基于隐马尔可夫模型的工业过程监测方法研", 《中国优秀博硕士学位论文全文数据库(博士) 信息科技辑》* |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113494428A (en)* | 2020-03-20 | 2021-10-12 | 新疆金风科技股份有限公司 | Fault detection method and device of wind generating set |
| CN113723546A (en)* | 2021-09-03 | 2021-11-30 | 江苏理工学院 | Bearing fault detection method and system based on discrete hidden Markov model |
| CN113723546B (en)* | 2021-09-03 | 2023-12-22 | 江苏理工学院 | Bearing fault detection method and system based on discrete hidden Markov model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111582392B (en) | An on-line monitoring method for multi-condition health status of key components of wind turbines | |
| CN107341349B (en) | Method, system, memory and controller for wind turbine health assessment | |
| CN108647786B (en) | Rotary machine on-line fault monitoring method based on deep convolution countermeasure neural network | |
| Shi et al. | Study of wind turbine fault diagnosis and early warning based on SCADA data | |
| CN118897956A (en) | A method for intelligently monitoring the operating status of hydropower plant equipment | |
| CN116025529B (en) | Autonomous health assessment method and self-healing regulation and control method and system for wind turbine generator | |
| CN107728059B (en) | Pitch system state evaluation method | |
| CN104005917B (en) | Method and system fan condition being predicted based on Bayesian inference mode | |
| CN110361193A (en) | Method for distinguishing is known for wind generating set pitch control bearing fault | |
| CN111794921B (en) | Onshore wind turbine blade icing diagnosis method based on migration component analysis | |
| CN115326393B (en) | A wind turbine bearing fault diagnosis method based on temperature information | |
| CN114320773B (en) | Wind turbine generator system fault early warning method based on power curve analysis and neural network | |
| CN113048012A (en) | Wind turbine generator yaw angle identification method and device based on Gaussian mixture model | |
| CN114065842A (en) | Wind turbine generator fault self-diagnosis method and system | |
| CN110705581A (en) | Pitch bearing fault identification method based on improved hidden Markov model | |
| CN110273818B (en) | A method for monitoring the icing fault of fan blades based on shaft transformation thickness classification | |
| CN119004157B (en) | Wind turbine generator converter temperature abnormality early warning method and device and electronic equipment | |
| Liu et al. | Twin prototype networks with noisy label self-correction for fault diagnosis of wind turbine gearboxes | |
| CN115985072A (en) | Wind turbine cabin temperature monitoring and early warning method and system based on machine learning | |
| CN119250292A (en) | Offshore wind turbine fault prediction method and device considering environmental factors | |
| Ahilan et al. | A self-improved optimizer-based CNN for wind turbine fault detection | |
| CN113822344A (en) | Wind turbine generator front bearing state monitoring method based on data driving | |
| CN112699598A (en) | Intelligent diagnosis method and device for abnormal oil temperature of gear box | |
| Peng et al. | Wind turbine blades icing failure prognosis based on balanced data and improved entropy | |
| Souza et al. | Evaluation of data based normal behavior models for fault detection in wind turbines |
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 | ||
| RJ01 | Rejection of invention patent application after publication | Application publication date:20200117 |