CN107977508A - A kind of dynamo bearing failure prediction method - Google Patents
A kind of dynamo bearing failure prediction method Download PDFInfo
- Publication number
- CN107977508A CN107977508A CN201711222655.3A CN201711222655A CN107977508A CN 107977508 A CN107977508 A CN 107977508A CN 201711222655 A CN201711222655 A CN 201711222655A CN 107977508 A CN107977508 A CN 107977508A
- Authority
- CN
- China
- Prior art keywords
- data
- residual error
- prediction
- dynamo bearing
- actual
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/17—Mechanical parametric or variational design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/30—Circuit design
- G06F30/32—Circuit design at the digital level
- G06F30/333—Design for testability [DFT], e.g. scan chain or built-in self-test [BIST]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- 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
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- General Physics & Mathematics (AREA)
- Computer Hardware Design (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Pure & Applied Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Wind Motors (AREA)
Abstract
The invention discloses a kind of dynamo bearing failure prediction method, the Wind turbines SCADA data that power curve method chooses health is primarily based on, as training sample and test sample；Then the dynamo bearing failure predication model of target component is established；Secondly, using the target component of prediction model prediction test sample, and and actual comparison, obtain residual error；Then process control technology is utilized, calculates the lower limit of the upper limit of alarm threshold value and the upper limit and warning threshold of the lower limit of alarm threshold value and warning threshold；Finally, using prediction model prediction Wind turbines actual motion target component, and and actual comparison, residual error one is obtained, and judge the health status of dynamo bearing.Beneficial effects of the present invention：Prediction is modeled using Wind turbines SCADA data, using statistical Process Control thought, the state of dynamo bearing is divided into three health, inferior health and failure states, the dynamo bearing that gives warning in advance fails, and improves the use managerial ability of bearing.
Description
Technical field
The present invention relates to failure prediction method technical field, it particularly relates to a kind of dynamo bearing failure predication side
Method.
Background technology
Wind turbines dynamo bearing is the indispensable parts of rotating machinery, and long-term work is easy in the presence of a harsh environment
Cause the Internal and external cycle, rolling element and retainer of bearing easily to break down or fail, influence the normal operation of whole unit, cause
Economic loss.The fault diagnosis for bearing and early warning carry out feature extraction and analysis primarily directed to CMS vibration signals at present,
But CMS monitoring devices are not that every Wind turbines have an installation, thus for bearing fault it is timely find and processing be have it is tired
Difficult.
The content of the invention
For the above-mentioned technical problem in correlation technique, the present invention proposes a kind of dynamo bearing failure prediction method, energy
Enough overcome the above-mentioned deficiency of the prior art.
To realize above-mentioned technical purpose, the technical proposal of the invention is realized in this way：
A kind of dynamo bearing failure prediction method, comprises the following steps：
S1：The Wind turbines SCADA data of health is chosen based on power curve method, as training sample and test specimens
This；
S2：Establish the dynamo bearing failure predication model of target component；
S3：Using the target component of prediction model prediction test sample, and and actual comparison, obtain residual error；
S4：Using process control technology, the upper of the upper limit of alarm threshold value and the lower limit of alarm threshold value and warning threshold is calculated
Limit and the lower limit of warning threshold；
S5：Using prediction model prediction Wind turbines actual motion target component, and and actual comparison, obtain residual error
One；
S6：If residual error one is more than warning lower limit and less than the warning upper limit, dynamo bearing health is judged；
S7：If residual error one be more than the warning threshold upper limit and less than the alarm threshold value upper limit or residual error be less than warning value lower limit and
More than alarm threshold value lower limit, then judge that dynamo bearing is in sub-health state；
S8：If residual error one is more than high alarm setting or less than low alarm setting, judge that dynamo bearing is in malfunction.
Further, in step sl, the Wind turbines SCADA data that health is chosen based on power curve method
Specific steps include：
S1.1：Reject and shut down data：More than incision wind speed and it is less than cut-out wind speed, power is less than or equal to zero data；
S1.2：Rejecting abnormalities data：Active power is less than -10KW or active power exceedes the number for completely sending out power 100kw
According to and wind speed exceed cut-out wind speed data；
S1.3：According to the actual power curve data of Wind turbines, using look-up table, it is corresponding to calculate each air speed data
Theoretical power (horse-power) value PIt is theoretical；
S1.4：Calculate theoretical performance number PIt is theoreticalWith actual power value PIt is actualAbsolute difference Err, i.e. Err=| PIt is actual-PIt is theoretical|；
S1.5：The data that ERR is more than to setting worst error 25KW are rejected；
S1.6：The abnormal data in more than 1 year Wind turbines SCADA data, remaining data are rejected using the above method
As Wind turbines health sample data.
Further, in step s 2, the dynamo bearing failure predication model i.e. BP for the establishing target component god
Through network model, specifically include:
S2.1：It is 15 state parameters to determine mode input, and model output is TOutput, i.e. generator drive end bearing temperature
With generator non-driven-end bearing temperature difference；
S2.2：Three layers of BP neural network are chosen as prediction model；
S2.3：According to kolmogorov theorems, the approximation relation for determining three layers of BP neural network the number of hidden nodes is：n2=
2n1+ 1,
Wherein：n1For input layer number；n2For hidden layer neuron number；
S2.4：The transmission function of bipolarity S type functions and linear function as model hidden layer and output layer is chosen respectively；
S2.5：Choose the weights of Bayesian regularization iterative algorithm Optimized BP Neural Network；
S2.6：Model is established using the function newff of the Neural Network Toolbox of MATLAB, by repeatedly training BP nerves
Network, obtains optimal model.
Preferably, in step S2.5, the performance function of the Bayesian regularization iterative algorithm is：
Msereg=α msw+ β mse
In formula：Mse is mean square error, and n is total sample number, tiFor desired output, piFor network reality output, msw is network
Weights are square, and m is total for network weight, ωjFor network weight, α and β are regularization coefficient.
Further, in step s3, it is described using prediction model prediction test sample target component, and with reality
The specific steps that value contrast obtains residual error include：
S3.1：The model trained using step S2, is predicted test sample, obtains predicted value TPrediction；
S3.2：By predicted value TPredictionWith actual value TIt is actualContrast, calculates residual error Error, and specific formula is：
Error=TPrediction-TIt is actual。
Further, in step s 4, the upper limit and warning level that utilize process control technology, calculate alarm threshold value
The specific steps of the lower limit of value and the lower limit of the upper limit of warning threshold and warning threshold include：
S4.1：Verified using the Wind turbines data under normal operational condition, it was demonstrated that the residual error of sample obeys normal state
Distribution, determines the mean μ and variances sigma of sample residual2As normal distribution average and the estimate of variance, calculation expression is such as
Under：
S4.2：The residual error data calculated using step S3, calculates residual error mean μ and variances sigma2, formula is as follows：
S4.3：Determine centered on mean μ, the alarm and warning of ± 3 σ and ± 2 σ as dynamo bearing fault pre-alarming
Bound：I.e. high alarm setting is the σ of μ+3；The warning upper limit is the σ of μ+2；Warning lower limit is μ -2 σ；Low alarm setting is μ -3 σ.
Further, in step s 5, the utilization prediction model prediction Wind turbines actual motion target component, and
The specific steps of residual error one are obtained with actual comparison to be included：
S5.1：The model trained using step S2, is predicted actual Wind turbines data, obtains predicted value
TPrediction 1；
S5.2：By predicted value TPrediction 1With actual value TActual 1Contrast, calculates one Error1 of residual error, and specific formula is：Error1=
TPrediction 1-TActual 1。
Beneficial effects of the present invention：The present invention can utilize Wind turbines on the basis of any hardware device is not increased
SCADA data is modeled prediction, and using statistical Process Control thought, the state of dynamo bearing is divided into health, inferior health
With three states of failure, the dynamo bearing that gives warning in advance fails, and improves the use managerial ability of bearing.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to institute in embodiment
Attached drawing to be used is needed to be briefly described, it should be apparent that, drawings in the following description are only some implementations of the present invention
Example, for those of ordinary skill in the art, without creative efforts, can also obtain according to these attached drawings
Obtain other attached drawings.
Fig. 1 is a kind of flow chart of the dynamo bearing failure prediction method described according to embodiments of the present invention；
Fig. 2 is the data prediction flow chart described according to embodiments of the present invention；
Fig. 3 is the BP neural network modeling procedure figure described according to embodiments of the present invention.
Embodiment
Below in conjunction with the attached drawing in the embodiment of the present invention, the technical solution in the embodiment of the present invention is carried out clear, complete
Site preparation describes, it is clear that described embodiment is only part of the embodiment of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, those of ordinary skill in the art's all other embodiments obtained, belong to what the present invention protected
Scope.
As shown in Figs. 1-3, a kind of dynamo bearing failure prediction method according to embodiments of the present invention, including it is following
Step：
S1：The Wind turbines SCADA data of health is chosen based on power curve method, as training sample and test specimens
This；
S2：Establish the dynamo bearing failure predication model of target component；
S3：Using the target component of prediction model prediction test sample, and and actual comparison, obtain residual error；
S4：Using process control technology, the upper of the upper limit of alarm threshold value and the lower limit of alarm threshold value and warning threshold is calculated
Limit and the lower limit of warning threshold；
S5：Using prediction model prediction Wind turbines actual motion target component, and and actual comparison, obtain residual error
One；
S6：If residual error one is more than warning lower limit and less than the warning upper limit, dynamo bearing health is judged；
S7：If residual error one be more than the warning threshold upper limit and less than the alarm threshold value upper limit or residual error be less than warning value lower limit and
More than alarm threshold value lower limit, then judge that dynamo bearing is in sub-health state；
S8：If residual error one is more than high alarm setting or less than low alarm setting, judge that dynamo bearing is in malfunction.
In one embodiment, in step sl, the Wind turbines that health is chosen based on power curve method
The specific steps of SCADA data include：
S1.1：Reject and shut down data：More than incision wind speed and it is less than cut-out wind speed, power is less than or equal to zero data；
S1.2：Rejecting abnormalities data：Active power is less than -10KW or active power exceedes the number for completely sending out power 100kw
According to and wind speed exceed cut-out wind speed data；
S1.3：According to the actual power curve data of Wind turbines, using look-up table, it is corresponding to calculate each air speed data
Theoretical power (horse-power) value PIt is theoretical；
S1.4：Calculate theoretical performance number PIt is theoreticalWith actual power value PIt is actualAbsolute difference Err, i.e. Err=| PIt is actual-PIt is theoretical|；
S1.5：The data that ERR is more than to setting worst error 25KW are rejected；
S1.6：The abnormal data in more than 1 year Wind turbines SCADA data, remaining data are rejected using the above method
As Wind turbines health sample data.
In one embodiment, in step s 2, the dynamo bearing failure predication mould for establishing target component
Type, that is, BP neural network model, specifically includes:
S2.1：It is 15 state parameters to determine mode input, and model output is TOutput, i.e. generator drive end bearing temperature
With generator non-driven-end bearing temperature difference；
S2.2：Three layers of BP neural network are chosen as prediction model；
S2.3：According to kolmogorov theorems, the approximation relation for determining three layers of BP neural network the number of hidden nodes is：n2=
2n1+ 1,
Wherein：n1For input layer number；n2For hidden layer neuron number；
S2.4：The transmission function of bipolarity S type functions and linear function as model hidden layer and output layer is chosen respectively；
S2.5：Choose the weights of Bayesian regularization iterative algorithm Optimized BP Neural Network；
S2.6：Model is established using the function newff of the Neural Network Toolbox of MATLAB, by repeatedly training BP nerves
Network, obtains optimal model.
In one embodiment, in step S2.5, the performance function of the Bayesian regularization iterative algorithm is：
Msereg=α msw+ β mse
In formula：Mse is mean square error, and n is total sample number, tiFor desired output, piFor network reality output, msw is network
Weights are square, and m is total for network weight, ωjFor network weight, α and β are regularization coefficient.
In one embodiment, in step s3, the target component using prediction model prediction test sample,
And the specific steps that residual error is obtained with actual comparison include：
S3.1：The model trained using step S2, is predicted test sample, obtains predicted value TPrediction；
S3.2：By predicted value TPredictionWith actual value TIt is actualContrast, calculates residual error Error, and specific formula is：
Error=TPrediction-TIt is actual。
In one embodiment, in step s 4, the upper limit for utilizing process control technology, calculating alarm threshold value
Include with the specific steps of the lower limit of alarm threshold value and the lower limit of the upper limit of warning threshold and warning threshold：
S4.1：Verified using the Wind turbines data under normal operational condition, it was demonstrated that the residual error of sample obeys normal state
Distribution, determines the mean μ and variances sigma of sample residual2As normal distribution average and the estimate of variance, calculation expression is such as
Under：
S4.2：The residual error data calculated using step S3, calculates residual error mean μ and variances sigma2, formula is as follows：
S4.3：Determine centered on mean μ, the alarm and warning of ± 3 σ and ± 2 σ as dynamo bearing fault pre-alarming
Bound：I.e. high alarm setting is the σ of μ+3；The warning upper limit is the σ of μ+2；Warning lower limit is μ -2 σ；Low alarm setting is μ -3 σ.
In one embodiment, in step s 5, the utilization prediction model prediction Wind turbines actual motion mesh
Parameter is marked, and the specific steps for obtaining with actual comparison residual error one include：
S5.1：The model trained using step S2, is predicted actual Wind turbines data, obtains predicted value
TPrediction 1；
S5.2：By predicted value TPrediction 1With actual value TActual 1Contrast, calculates one Error1 of residual error, and specific formula is：Error1=
TPrediction 1-TActual 1。
In order to facilitate understand the present invention above-mentioned technical proposal, below by way of in specifically used mode to the present invention it is above-mentioned
Technical solution is described in detail.
When specifically used, data are pre-processed as shown in Fig. 2, being primarily based on, extract health data sample, specifically
To reject shutdown data first, that is, it is more than incision wind speed and is less than or equal to zero data less than cut-out wind speed and power；Then pick
Except abnormal data, i.e., active power is less than -10KW or active power exceedes the data for completely sending out power 100kw and wind speed beyond cutting
Go out the data of wind speed；Secondly, according to the actual power curve data of Wind turbines, using look-up table, each air speed data is calculated
Corresponding theoretical power (horse-power) value PIt is theoretical；Then theoretical performance number P is calculatedIt is theoreticalWith actual power value PIt is actualAbsolute difference Err, i.e. Err=
|PIt is actual-PIt is theoretical|；The data that ERR is more than to setting worst error 25KW again are rejected；Finally, using the above method reject 1 year with
Abnormal data in upper Wind turbines SCADA (data acquisition and supervisor control) data, remaining data are as wind turbine
The healthy sample data of group；The number of data of accounting 75% is randomly selected from sample data (per data by 15 state parameters
Composition) training data is used as, residue 25% is used as test data, generator drive end bearing temperature and generator anti-drive end axis
Temperature gap is held as target component.
Then, the dynamo bearing failure predication model of target component, i.e. BP neural network model are established, it is specific as schemed
Shown in 3, it is first determined mode input is 15 state parameters, and model output is TOutput, i.e., generator drive end bearing temperature is with sending out
Motor non-driven-end bearing temperature difference；Then three layers of BP neural network (i.e. input layer, hidden layer, output layer) are chosen as prediction
Model；Secondly, according to kolmogorov (Ke Ermoge loves) theorem, the approximation of three layers of BP neural network the number of hidden nodes is determined
Relation is：n2=2n1+ 1, wherein：n1For input layer number；n2For hidden layer neuron number；Then choose respectively double
The transmission function of polarity S type function and linear function as model hidden layer and output layer；Again, Bayesian regularization is chosen to change
For the weights of algorithm optimization BP neural network, wherein, in Bayesian regularization iterative algorithm, performance function is：
Msereg=α msw+ β mse
In formula：Mse is mean square error, and n is total sample number, tiFor desired output, piFor network reality output, msw is network
Weights are square, and m is total for network weight, ωjFor network weight, α and β are regularization coefficient；Finally, the nerve of MATLAB is utilized
The function newff of network tool case establishes model, by repeatedly training BP neural network, obtains optimal model.
Secondly, using the target component of prediction model prediction test sample, and and actual comparison, obtain residual error；Specifically
For the model trained first with second step, is predicted test sample, obtains predicted value TPrediction；Then by predicted value
TPredictionWith actual value TIt is actualContrast, calculates residual error Error, and specific formula is：Error=TPrediction-TIt is actual。
Then, using process control technology, the upper limit of alarm threshold value and the lower limit of alarm threshold value and warning threshold are calculated
The lower limit of the upper limit and warning threshold；Wind turbines generator drive end and non-driven-end bearing temperature difference prediction mould under normal condition
After type is established, Wind turbines generator drive end temperature and anti-drive end temperature gap T can be predicted using the modelOutput, in advance
Residual values between measured value and actual value always fluctuate within the specific limits, even if accidentally there is relatively large deviation, but not are formed and
Gesture；Once residual error amplitude persistently occurs increasing and exceedes tolerance interval, then it is assumed that there are the fortune that abnormal factors influence generator
OK, dynamo bearing has been likely to occur exception.
Specifically, verified first with the Wind turbines data under normal operational condition, it was demonstrated that the residual error clothes of sample
From normal distribution, the mean μ and variances sigma of sample residual are determined2As normal distribution average and the estimate of variance, calculation expression
Formula is as follows：
Then the residual error data calculated using the 3rd step, calculates residual error mean μ and variances sigma2, formula is as follows：
Secondly, determine centered on mean μ, the alarm and warning of ± 3 σ and ± 2 σ as dynamo bearing fault pre-alarming
Bound：I.e. high alarm setting is the σ of μ+3；The warning upper limit is the σ of μ+2；Warning lower limit is μ -2 σ；Low alarm setting is μ -3 σ.
Finally, predict Wind turbines actual motion target component using prediction model, and residual error is obtained with actual comparison
One, the model specially trained first with second step, is predicted actual Wind turbines data, obtains predicted value
TPrediction 1；Then by predicted value TPrediction 1With actual value TActual 1Contrast, calculates one Error1 of residual error, and specific formula is：Error1=
TPrediction 1-TActual 1, finally result is analyzed, if residual error one is more than warning lower limit and less than the warning upper limit, judges generator shaft
Hold health；If residual error one is more than the warning threshold upper limit and is less than warning value lower limit less than the alarm threshold value upper limit or residual error and is more than
Alarm threshold value lower limit, then judge that dynamo bearing is in sub-health state；If residual error one is more than high alarm setting or less than under alarm
Limit, then judge that dynamo bearing is in malfunction.
The present invention proposes a kind of dynamo bearing failure prediction method, predicts the healthy shape of Wind turbines dynamo bearing
State, give warning in advance problem, is conducive to wind field reasonable arrangement repair schedule.
In conclusion the present invention can utilize Wind turbines SCADA data on the basis of any hardware device is not increased
Prediction is modeled, using statistical Process Control thought, the state of dynamo bearing is divided into health, inferior health and failure three
State, the dynamo bearing that gives warning in advance fail, and improve the use managerial ability of bearing.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
With within principle, any modification, equivalent replacement, improvement and so on, should all be included in the protection scope of the present invention god.
Claims (7)
1. a kind of dynamo bearing failure prediction method, it is characterised in that comprise the following steps：
S1：The Wind turbines SCADA data of health is chosen based on power curve method, as training sample and test sample；
S2：Establish the dynamo bearing failure predication model of target component；
S3：Using the target component of prediction model prediction test sample, and and actual comparison, obtain residual error；
S4：Using process control technology, calculate the upper limit of alarm threshold value and the upper limit of the lower limit of alarm threshold value and warning threshold with
The lower limit of warning threshold；
S5：Using prediction model prediction Wind turbines actual motion target component, and and actual comparison, obtain residual error one；
S6：If residual error one is more than warning lower limit and less than the warning upper limit, dynamo bearing health is judged；
S7：If residual error one is more than the warning threshold upper limit and is less than warning value lower limit less than the alarm threshold value upper limit or residual error and is more than
Alarm threshold value lower limit, then judge that dynamo bearing is in sub-health state；
S8：If residual error one is more than high alarm setting or less than low alarm setting, judge that dynamo bearing is in malfunction.
A kind of 2. dynamo bearing failure prediction method according to claim 1, it is characterised in that in step sl, institute
The specific steps for the Wind turbines SCADA data for choosing health based on power curve method stated include：
S1.1：Reject and shut down data：More than incision wind speed and it is less than cut-out wind speed, power is less than or equal to zero data；
S1.2：Rejecting abnormalities data：Active power be less than -10KW or active power exceed the data of completely sending out power 100kw and
Wind speed exceeds the data of cut-out wind speed；
S1.3：According to the actual power curve data of Wind turbines, using look-up table, the corresponding theory of each air speed data is calculated
Performance number；
S1.4：Calculate theoretical performance numberWith actual power valueAbsolute difference Err, i.e.,；
S1.5：The data that ERR is more than to setting worst error 25KW are rejected；
S1.6：The abnormal data in more than 1 year Wind turbines SCADA data, remaining data conduct are rejected using the above method
Wind turbines health sample data.
A kind of 3. dynamo bearing failure prediction method according to claim 1, it is characterised in that in step s 2, institute
That states establishes the dynamo bearing failure predication model i.e. BP neural network model of target component, specifically includes:
S2.1：It is 15 state parameters to determine mode input, and model output is, i.e., generator drive end bearing temperature is with sending out
Motor non-driven-end bearing temperature difference；
S2.2：Three layers of BP neural network are chosen as prediction model；
S2.3：According to kolmogorov theorems, the approximation relation for determining three layers of BP neural network the number of hidden nodes is：n2=2n1+ 1,
Wherein：n1For input layer number；n2For hidden layer neuron number；
S2.4：The transmission function of bipolarity S type functions and linear function as model hidden layer and output layer is chosen respectively；
S2.5：Choose the weights of Bayesian regularization iterative algorithm Optimized BP Neural Network；
S2.6：Model is established using the function newff of the Neural Network Toolbox of MATLAB, by repeatedly training BP nerve nets
Network, obtains optimal model.
A kind of 4. dynamo bearing failure prediction method according to claim 3, it is characterised in that in step S2.5,
The performance function of the Bayesian regularization iterative algorithm is：
In formula：For mean square error,For total sample number,For desired output,For network reality output,For network
Weights are square,It is total for network weight,For network weight,WithFor regularization coefficient.
A kind of 5. dynamo bearing failure prediction method according to claim 1, it is characterised in that in step s3, institute
The target component using prediction model prediction test sample stated, and the specific steps for obtaining with actual comparison residual error include：
S3.1：The model trained using step S2, is predicted test sample, obtains predicted value；
S3.2：By predicted valueWith actual valueContrast, calculates residual error, specific formula is：。
A kind of 6. dynamo bearing failure prediction method according to claim 1, it is characterised in that in step s 4, institute
That states utilizes process control technology, calculates the upper limit of alarm threshold value and the upper limit and warning of the lower limit of alarm threshold value and warning threshold
The specific steps of the lower limit of threshold value include：
S4.1：Verified using the Wind turbines data under normal operational condition, it was demonstrated that the residual error Normal Distribution of sample,
Determine the average of sample residualAnd varianceIt is as follows as normal distribution average and the estimate of variance, calculation expression：
；
S4.2：The residual error data calculated using step S3, calculates residual error averageAnd variance, formula is as follows：
；
S4.3：Determine with averageCentered on,WithAs dynamo bearing fault pre-alarming alarm and warning up and down
Limit：I.e. high alarm setting is；Alerting the upper limit is；Alerting lower limit is；Low alarm setting is。
A kind of 7. dynamo bearing failure prediction method according to claim 1, it is characterised in that in step s 5, institute
That states obtains the specific step of residual error one using prediction model prediction Wind turbines actual motion target component, and with actual comparison
Suddenly include：
S5.1：The model trained using step S2, is predicted actual Wind turbines data, obtains predicted value；
S5.2：By predicted valueWith actual valueContrast, calculates residual error one1, specific formula is：。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711222655.3A CN107977508A (en) | 2017-11-29 | 2017-11-29 | A kind of dynamo bearing failure prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711222655.3A CN107977508A (en) | 2017-11-29 | 2017-11-29 | A kind of dynamo bearing failure prediction method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107977508A true CN107977508A (en) | 2018-05-01 |
Family
ID=62008482
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711222655.3A Pending CN107977508A (en) | 2017-11-29 | 2017-11-29 | A kind of dynamo bearing failure prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107977508A (en) |
Cited By (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109118384A (en) * | 2018-07-16 | 2019-01-01 | 湖南优利泰克自动化系统有限公司 | A kind of Wind turbines healthy early warning method |
| CN109192340A (en) * | 2018-07-23 | 2019-01-11 | 广东核电合营有限公司 | The method and device of kilowatt pressurized water reactor nuclear power station generator measurement point monitoring |
| CN109307553A (en) * | 2018-09-27 | 2019-02-05 | 潍柴重机股份有限公司 | Base bearing based on base bearing temperature deviation overhauls reminding method |
| CN109670145A (en) * | 2018-12-12 | 2019-04-23 | 中国航空工业集团公司上海航空测控技术研究所 | Helicopter Main Reducer vibration performance threshold setting method based on observer |
| CN109829245A (en) * | 2019-02-25 | 2019-05-31 | 中科诺维(北京)科技有限公司 | Bearing fault method for early warning and device |
| CN110108457A (en) * | 2019-04-18 | 2019-08-09 | 大唐东北电力试验研究院有限公司 | A kind of First air machine shaft temperature diagnostic method based on gravitation neural network |
| CN110135064A (en) * | 2019-05-15 | 2019-08-16 | 上海交通大学 | A kind of generator rear bearing temperature fault prediction technique, system and controller |
| CN110263433A (en) * | 2019-06-19 | 2019-09-20 | 苏州科技大学 | A kind of fuse failure alarm method and system |
| CN110414155A (en) * | 2019-07-31 | 2019-11-05 | 北京天泽智云科技有限公司 | A kind of detection of fan part temperature anomaly and alarm method with single measuring point |
| CN110414154A (en) * | 2019-07-31 | 2019-11-05 | 北京天泽智云科技有限公司 | A kind of detection of fan part temperature anomaly and alarm method with double measuring points |
| CN110502770A (en) * | 2018-10-11 | 2019-11-26 | 起动科技有限公司 | Method, apparatus, equipment and storage medium based on ANN prediction mechanics parameter |
| CN110992205A (en) * | 2019-11-28 | 2020-04-10 | 中国船舶重工集团海装风电股份有限公司 | State detection method and system for generator winding of wind turbine generator and related components |
| CN110991666A (en) * | 2019-11-25 | 2020-04-10 | 远景智能国际私人投资有限公司 | Fault detection method, model training method, device, equipment and storage medium |
| CN111103136A (en) * | 2019-12-13 | 2020-05-05 | 南京松数科技有限公司 | Fan gearbox fault detection method based on SCADA data analysis |
| CN111275331A (en) * | 2020-01-20 | 2020-06-12 | 张家口卷烟厂有限责任公司 | Monitoring method and device for production system |
| CN111396266A (en) * | 2020-03-05 | 2020-07-10 | 浙江运达风电股份有限公司 | GBRT-based wind turbine generator bearing fault early warning method |
| CN111611718A (en) * | 2020-05-27 | 2020-09-01 | 江南大学 | Actuator fault estimation method based on Bayesian learning |
| CN111723998A (en) * | 2020-06-28 | 2020-09-29 | 国网湖南省电力有限公司 | Early warning method for fault of oil pressure system of generator oil pressure tank |
| CN111814849A (en) * | 2020-06-22 | 2020-10-23 | 浙江大学 | DA-RNN-based wind turbine generator key component fault early warning method |
| CN112067289A (en) * | 2020-08-21 | 2020-12-11 | 天津电气科学研究院有限公司 | Motor shaft and transmission shaft abnormal vibration early warning algorithm based on neural network |
| CN112324627A (en) * | 2020-09-22 | 2021-02-05 | 大唐可再生能源试验研究院有限公司 | Wind generating set generator bearing temperature alarm system |
| CN112390127A (en) * | 2020-12-12 | 2021-02-23 | 中铁第四勘察设计院集团有限公司 | Health degree model-based preventive maintenance strategy generation method for escalator |
| CN112598172A (en) * | 2020-12-21 | 2021-04-02 | 北京华能新锐控制技术有限公司 | Wind turbine bearing temperature early warning method |
| CN112784373A (en) * | 2021-01-19 | 2021-05-11 | 河北大学 | Fault early warning method for wind turbine generator gearbox |
| CN113312851A (en) * | 2021-06-16 | 2021-08-27 | 华电山东新能源有限公司 | Early warning method for temperature abnormity of main bearing of wind driven generator |
| CN113468818A (en) * | 2021-07-13 | 2021-10-01 | 上海电气风电集团股份有限公司 | Method and device for monitoring winding running state of wind generating set |
| CN114037127A (en) * | 2021-10-25 | 2022-02-11 | 华能射阳新能源发电有限公司 | Fault prediction method and device for wind turbine generator |
| EP3696405B1 (en) | 2019-02-14 | 2023-03-22 | Mitsubishi Heavy Industries, Ltd. | Operating state evaluation method and operating state evaluation device |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107153929A (en) * | 2017-07-10 | 2017-09-12 | 龙源（北京）风电工程技术有限公司 | Gearbox of wind turbine fault monitoring method and system based on deep neural network |
-
2017
- 2017-11-29 CN CN201711222655.3A patent/CN107977508A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107153929A (en) * | 2017-07-10 | 2017-09-12 | 龙源（北京）风电工程技术有限公司 | Gearbox of wind turbine fault monitoring method and system based on deep neural network |
Non-Patent Citations (6)
| Title |
|---|
| 何永枢: "基于BP神经网络模型的故障预测分析", 《微计算机信息》 * |
| 孙俊: "《光谱技术在农作物/农产品信息无损检测中的应用》", 30 June 2017, 东南大学出版社 * |
| 张永辉: "基于监测数据的风力发电机故障预警研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
| 金晓航 等: "风力发电机组故障诊断与预测技术研究综述", 《仪器仪表学报》 * |
| 黄丽丽: "基于SCADA的风力机故障预测与健康管理技术研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
| 龙泉 等: "基于BP神经网络的多变量风力发电机绕组温度预测", 《2016年清洁高效发电技术协作网年会论文集》 * |
Cited By (36)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109118384A (en) * | 2018-07-16 | 2019-01-01 | 湖南优利泰克自动化系统有限公司 | A kind of Wind turbines healthy early warning method |
| CN109192340A (en) * | 2018-07-23 | 2019-01-11 | 广东核电合营有限公司 | The method and device of kilowatt pressurized water reactor nuclear power station generator measurement point monitoring |
| CN109307553A (en) * | 2018-09-27 | 2019-02-05 | 潍柴重机股份有限公司 | Base bearing based on base bearing temperature deviation overhauls reminding method |
| CN110502770A (en) * | 2018-10-11 | 2019-11-26 | 起动科技有限公司 | Method, apparatus, equipment and storage medium based on ANN prediction mechanics parameter |
| CN109670145A (en) * | 2018-12-12 | 2019-04-23 | 中国航空工业集团公司上海航空测控技术研究所 | Helicopter Main Reducer vibration performance threshold setting method based on observer |
| EP3696405B1 (en) | 2019-02-14 | 2023-03-22 | Mitsubishi Heavy Industries, Ltd. | Operating state evaluation method and operating state evaluation device |
| CN109829245A (en) * | 2019-02-25 | 2019-05-31 | 中科诺维(北京)科技有限公司 | Bearing fault method for early warning and device |
| CN110108457A (en) * | 2019-04-18 | 2019-08-09 | 大唐东北电力试验研究院有限公司 | A kind of First air machine shaft temperature diagnostic method based on gravitation neural network |
| CN110135064A (en) * | 2019-05-15 | 2019-08-16 | 上海交通大学 | A kind of generator rear bearing temperature fault prediction technique, system and controller |
| CN110135064B (en) * | 2019-05-15 | 2023-07-18 | 上海交通大学 | Method, system and controller for predicting temperature faults of rear bearing of generator |
| CN110263433B (en) * | 2019-06-19 | 2024-03-05 | 苏州科技大学 | Fuse fault alarm method and system |
| CN110263433A (en) * | 2019-06-19 | 2019-09-20 | 苏州科技大学 | A kind of fuse failure alarm method and system |
| CN110414155A (en) * | 2019-07-31 | 2019-11-05 | 北京天泽智云科技有限公司 | A kind of detection of fan part temperature anomaly and alarm method with single measuring point |
| CN110414155B (en) * | 2019-07-31 | 2022-09-30 | 北京天泽智云科技有限公司 | Fan component temperature abnormity detection and alarm method with single measuring point |
| CN110414154A (en) * | 2019-07-31 | 2019-11-05 | 北京天泽智云科技有限公司 | A kind of detection of fan part temperature anomaly and alarm method with double measuring points |
| CN110414154B (en) * | 2019-07-31 | 2022-09-30 | 北京天泽智云科技有限公司 | Fan component temperature abnormity detection and alarm method with double measuring points |
| CN110991666A (en) * | 2019-11-25 | 2020-04-10 | 远景智能国际私人投资有限公司 | Fault detection method, model training method, device, equipment and storage medium |
| CN110991666B (en) * | 2019-11-25 | 2023-09-15 | 远景智能国际私人投资有限公司 | Fault detection method, training device, training equipment and training equipment for model, and storage medium |
| CN110992205A (en) * | 2019-11-28 | 2020-04-10 | 中国船舶重工集团海装风电股份有限公司 | State detection method and system for generator winding of wind turbine generator and related components |
| CN111103136A (en) * | 2019-12-13 | 2020-05-05 | 南京松数科技有限公司 | Fan gearbox fault detection method based on SCADA data analysis |
| CN111275331A (en) * | 2020-01-20 | 2020-06-12 | 张家口卷烟厂有限责任公司 | Monitoring method and device for production system |
| CN111396266A (en) * | 2020-03-05 | 2020-07-10 | 浙江运达风电股份有限公司 | GBRT-based wind turbine generator bearing fault early warning method |
| CN111611718A (en) * | 2020-05-27 | 2020-09-01 | 江南大学 | Actuator fault estimation method based on Bayesian learning |
| CN111611718B (en) * | 2020-05-27 | 2023-04-07 | 江南大学 | Actuator fault estimation method based on Bayesian learning |
| CN111814849A (en) * | 2020-06-22 | 2020-10-23 | 浙江大学 | DA-RNN-based wind turbine generator key component fault early warning method |
| CN111814849B (en) * | 2020-06-22 | 2024-02-06 | 浙江大学 | DA-RNN-based wind turbine generator set key component fault early warning method |
| CN111723998A (en) * | 2020-06-28 | 2020-09-29 | 国网湖南省电力有限公司 | Early warning method for fault of oil pressure system of generator oil pressure tank |
| CN112067289A (en) * | 2020-08-21 | 2020-12-11 | 天津电气科学研究院有限公司 | Motor shaft and transmission shaft abnormal vibration early warning algorithm based on neural network |
| CN112324627A (en) * | 2020-09-22 | 2021-02-05 | 大唐可再生能源试验研究院有限公司 | Wind generating set generator bearing temperature alarm system |
| CN112390127A (en) * | 2020-12-12 | 2021-02-23 | 中铁第四勘察设计院集团有限公司 | Health degree model-based preventive maintenance strategy generation method for escalator |
| CN112598172A (en) * | 2020-12-21 | 2021-04-02 | 北京华能新锐控制技术有限公司 | Wind turbine bearing temperature early warning method |
| CN112784373B (en) * | 2021-01-19 | 2022-03-01 | 河北大学 | Fault early warning method for wind turbine generator gearbox |
| CN112784373A (en) * | 2021-01-19 | 2021-05-11 | 河北大学 | Fault early warning method for wind turbine generator gearbox |
| CN113312851A (en) * | 2021-06-16 | 2021-08-27 | 华电山东新能源有限公司 | Early warning method for temperature abnormity of main bearing of wind driven generator |
| CN113468818A (en) * | 2021-07-13 | 2021-10-01 | 上海电气风电集团股份有限公司 | Method and device for monitoring winding running state of wind generating set |
| CN114037127A (en) * | 2021-10-25 | 2022-02-11 | 华能射阳新能源发电有限公司 | Fault prediction method and device for wind turbine generator |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107977508A (en) | A kind of dynamo bearing failure prediction method | |
| Kamal et al. | Robust fuzzy fault-tolerant control of wind energy conversion systems subject to sensor faults | |
| Odgaard et al. | A benchmark evaluation of fault tolerant wind turbine control concepts | |
| Biswal et al. | Design and development of a wind turbine test rig for condition monitoring studies | |
| Li et al. | Fuzzy neural network technique for system state forecasting | |
| CN108647786A (en) | The rotating machinery on-line fault monitoring method of neural network is fought based on depth convolution | |
| Zhan et al. | Health assessment methods for wind turbines based on power prediction and mahalanobis distance | |
| Badihi et al. | Fault-tolerant cooperative control in a wind farm using adaptive control reconfiguration and control reallocation | |
| Badihi et al. | A review on application of monitoring, diagnosis, and fault-tolerant control to wind turbines | |
| Rahnavard et al. | On the development of a sliding mode observer-based fault diagnosis scheme for a wind turbine benchmark model | |
| Du et al. | A SCADA data based anomaly detection method for wind turbines | |
| Badihi et al. | Passive fault-tolerant cooperative control in an offshore wind farm | |
| Baltazar et al. | A review on neurocomputing based wind turbines fault diagnosis and prognosis | |
| Badihi et al. | Active fault tolerant control in a wind farm with decreased power generation due to blade erosion/debris build-up | |
| Sami et al. | Wind turbine sensor fault tolerant control via a multiple-model approach | |
| Zemali et al. | Fault diagnosis-based observers using Kalman filters and Luenberger estimators: Application to the pitch system fault actuators | |
| JP6246189B2 (en) | Information processing system and method for monitoring complex systems | |
| Odgaard et al. | Fault diagnosis and fault tolerant control with application on a wind turbine low speed shaft encoder | |
| Elyassami et al. | Bearing fault diagnosis and classification based on KDA and alpha-stable fusion | |
| Shi et al. | Research on fault classification of wind turbine based on IMF kurtosis and PSO-SOM-LVQ | |
| Fu et al. | Fault classification for wind turbine benchmark model based on hilbert-huang transformation and support vector machine strategies | |
| Elyassami et al. | Gear fault diagnosis and classification using data vibration | |
| Wu et al. | Machine learning approach for shaft crack detection through acoustical emission signals | |
| Dinh et al. | Implementation of Digital Twin-Assisted Condition Monitoring and Fault Diagnosis for Wind Turbines | |
| Talebi et al. | Fault detection of wind energy conversion systems using recurrent neural networks |
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: 20180501 |