CN115563501A - Wind turbine generator yaw stuck diagnosis method and system based on SVM algorithm - Google Patents
Wind turbine generator yaw stuck diagnosis method and system based on SVM algorithm Download PDFInfo
- Publication number
- CN115563501A CN115563501A CN202211197505.2A CN202211197505A CN115563501A CN 115563501 A CN115563501 A CN 115563501A CN 202211197505 A CN202211197505 A CN 202211197505A CN 115563501 A CN115563501 A CN 115563501A
- Authority
- CN
- China
- Prior art keywords
- yaw
- data
- wind turbine
- turbine generator
- stuck
- 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
- 238000003745 diagnosis Methods 0.000 title claims abstract description 80
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 37
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000012549 training Methods 0.000 claims abstract description 55
- 238000012360 testing method Methods 0.000 claims abstract description 32
- 230000000694 effects Effects 0.000 claims abstract description 17
- 238000002372 labelling Methods 0.000 claims abstract description 16
- 238000004458 analytical method Methods 0.000 claims abstract description 12
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 230000000903 blocking effect Effects 0.000 claims description 28
- 230000008859 change Effects 0.000 claims description 24
- 230000009471 action Effects 0.000 claims description 23
- 238000010606 normalization Methods 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 8
- 230000002159 abnormal effect Effects 0.000 claims description 7
- 238000000605 extraction Methods 0.000 claims description 5
- 238000012216 screening Methods 0.000 claims description 5
- 238000012795 verification Methods 0.000 claims description 5
- 238000012886 linear function Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- 238000012937 correction Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 2
- 238000012423 maintenance Methods 0.000 abstract description 10
- 230000008569 process Effects 0.000 abstract description 3
- 238000001514 detection method Methods 0.000 description 13
- 238000007405 data analysis Methods 0.000 description 5
- 238000011161 development Methods 0.000 description 3
- 238000010248 power generation Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000002405 diagnostic procedure Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- 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
- Y02E10/00—Energy generation through renewable energy sources
- Y02E10/70—Wind energy
- Y02E10/72—Wind turbines with rotation axis in wind direction
Landscapes
- Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Wind Motors (AREA)
Abstract
The invention discloses a wind turbine generator yaw stuck diagnosis method and system based on an SVM algorithm, wherein the method comprises the following steps of: s1, acquiring original data; s2, preprocessing the original data; s3, extracting features; s4, dividing the training data into a training set and a test set, and labeling the training set; s5, inputting the marked training set into an SVM algorithm for training to obtain a yaw stuck diagnosis model; s6, calculating the accuracy of the test set and constructing an ROC curve; s7, evaluating the classification effect of the yaw stuck diagnosis model; s8, manually checking the accuracy of the actual identification result of the yaw stuck diagnosis model, and if the actual identification result is qualified, issuing the yaw stuck diagnosis model and starting analysis; according to the invention, by acquiring the sensor data of the wind turbine generator and improving and optimizing the yaw control algorithm, the yaw jamming in the operation process of the wind turbine generator can be rapidly and effectively identified, early warning is timely given back to the site, the component failure rate and the operation and maintenance cost are reduced, and the cost is saved.
Description
Technical Field
The invention relates to the technical field of wind turbine generator yaw diagnosis, in particular to a wind turbine generator yaw stuck diagnosis method and system based on an SVM algorithm.
Background
The wind turbine generator runs in a severe natural environment for a long time, although the wind turbine generator considers limit load caused by extreme wind conditions during research, development and design, the practical situation is inconsistent with the simulation situation due to deviation between actual wind resources and data capable of being researched, limit working conditions can still occur during running, the design limit is exceeded, damage to all parts of the wind turbine generator is caused, and finally faults of the wind turbine generator occur frequently. The yawing system of the wind driven generator plays an important role in generating capacity, the frequent yawing and blocking are found in early stage and are repaired or the system is adjusted and optimized in a targeted manner, and compared with the condition that the part is damaged due to the development of a fault, the system needs to be stopped and replaced, so that the economic loss is low. High-frequency yaw blocking can cause yaw driving to be damaged and further cause yaw system faults, so that the wind turbine needs to be subjected to unplanned shutdown maintenance, the power generation performance is reduced slightly, key components such as a yaw motor and a yaw brake in the yaw system are seriously damaged, in addition, the fatigue load of the wind turbine can be increased indirectly through vibration caused by the yaw blocking, the influence causes the wind turbine to increase the operation and maintenance cost and pressure for a wind turbine manufacturer, the operation performance of the wind turbine is also deteriorated, and serious economic loss is brought. Therefore, the rapid detection of the yaw seizure of the wind turbine generator and the early warning become the most important concern in the wind power generation industry, but at the present stage, the research on the yaw seizure detection is still few.
The existing wind turbine generator fault detection method mainly comprises a data-based fault detection method and a model-based fault detection method, wherein the data-based fault detection method avoids a complex time-consuming modeling process, can quickly detect faults, but has the limit of identifying the faults through manually setting a relevant threshold or manually observing, is easy to cause misjudgment when a system is disturbed or the state in the operation process is changed, and can mine a rule through manual data analysis when massive fault data are faced, so that the efficiency is far from expectation; the later can fully utilize the internal deep information of the whole machine set system, effectively reflect the essential characteristics of physical system faults, not only solve the problem of complicated manual calculation amount, but also can mine the rules which cannot be identified by manpower and are hidden, carry out self-learning by data driving, and continuously improve the accuracy rate of the early warning model.
With the rapid development of artificial intelligence technology, machine learning and deep learning methods are widely applied in the field of wind power, for example, a neural network is used for fault detection of a variable pitch system of a wind turbine generator; based on an improved random forest algorithm, fault detection and early warning are carried out on large parts of the wind turbine generator, but the method is not applied to the yaw blocking problem of the wind turbine generator.
Disclosure of Invention
The invention aims to solve the defects in the prior art and provides a yaw jamming diagnosis method of a wind turbine generator based on an SVM algorithm.
The invention provides a wind turbine generator yaw stuck diagnosis system based on an SVM algorithm.
The first purpose of the invention is realized by the following technical scheme: a wind turbine generator yaw stuck diagnosis method based on an SVM algorithm comprises the following steps:
s1, acquiring sensor data recording fault actions of a wind turbine generator as original data;
s2, carrying out normalization pretreatment on the original data;
s3, extracting the characteristics of the preprocessed original data to obtain model data;
s4, performing up-sampling processing on the model data, dividing the model data into a training set and a test set, and labeling the training set according to a manual labeling and weak supervision learning method;
s5, inputting the marked training set into an SVM algorithm for training to obtain a yaw stuck diagnosis model;
s6, verifying and calculating the accuracy of the test set, and constructing an ROC curve according to the normal yaw and yaw blockage data of the wind turbine generator;
s7, evaluating the effect of the yaw stuck diagnosis model according to the accuracy of the test set and the ROC curve, and executing the step S8 if the evaluated effect of the yaw stuck diagnosis model meets the preset requirement; if the effect of the estimated yaw stuck diagnosis model does not meet the preset requirement, repeating the steps S1 to S6 until the effect of the estimated yaw stuck diagnosis model meets the preset requirement, and executing a step S8;
s8, manually checking the accuracy of the actual recognition result of the yaw stuck diagnosis model, if the actual recognition result is unqualified, adding a new data sample into a training set, and repeating the steps from S5 to S7 until the actual recognition result is qualified; and if the yaw blockage diagnosis model is qualified, issuing the yaw blockage diagnosis model and starting analysis so as to early warn the yaw fault problem of the wind turbine generator.
Further, the step S1 includes the steps of:
the sensor data are Tracelog data, and each Tracelog data is millisecond-level data in a time period of one minute before and after a fault action of the wind turbine generator;
screening is carried out on the obtained Tracelog data, and data recording time periods of yawing and stopping actions of the wind turbine generator are extracted: comparing the change rate of the azimuth angles of the engine room, when the wind turbine generator is in normal yaw, the change rate of the azimuth angles of the engine room is constant, namely the change rates of the azimuth angles of the engine room at adjacent moments are equal, and the change rates of the azimuth angles of the engine room at adjacent moments are unequal when the wind turbine generator is in abnormal yaw, and at the moment, the wind turbine generator is regarded as yaw blockage.
Further, the step S2 includes the steps of:
performing linear transformation on the original data by adopting linear function normalization:
the original data comprises continuous value variables and classification variables, the continuous value variables comprise the maximum value of the variation of the azimuth angle of the engine room, the mean value of the variation of the azimuth angle of the engine room, the quartile of the variation of the azimuth angle of the engine room, the mean value of wind speed, the standard deviation of wind direction and the variety of the variation of the azimuth angle of the engine room, and the classification variables comprise the yaw direction, the electromagnetic band-type brake and the grid-connected mode;
and mapping the continuous value variables to the range of [0,1] to realize the equal ratio scaling of each continuous value variable, wherein the normalization formula is as follows:
wherein X is the value of each continuous variable, X min Is the minimum of the values of the corresponding continuous-valued variables, X max Is the maximum value of the numerical values of the corresponding continuous value variables;
meanwhile, the yaw direction is coded, the clockwise yaw CW is recorded as 0, the counterclockwise yaw CCW is recorded as 1, and the non-yaw state is recorded as 2.
Further, the step S3 includes the steps of:
and extracting the type of azimuth angle variation of the yaw opportunity cabin, the mean value of the azimuth angle variation of the yaw opportunity cabin, the standard deviation of the azimuth angle variation of the yaw opportunity cabin, the maximum value of the azimuth angle variation of the yaw opportunity cabin, the quartile of the azimuth angle variation of the cabin, the yaw direction, the average wind speed value and the standard wind direction difference from the preprocessed original data as characteristic values.
Further, in step S6, the constructing an ROC curve according to the normal yaw and yaw stuck data of the wind turbine includes the following steps:
evaluating the classification capability of the yaw Kadun diagnosis model through an area AUC value calculated by an ROC curve, wherein the abscissa of the ROC curve is a false positive rate FPR, and the ordinate is a true positive rate TPR; the formula for FPR and TPR is as follows:
wherein, P is the number of the yaw stuck data, N is the number of the normal yaw data, TP is the number of the yaw stuck data predicted by the yaw stuck diagnostic model in the P yaw stuck data, and FP is the number of the normal yaw predicted by the yaw stuck diagnostic model in the N normal yaw data.
Further, the step S7 includes the steps of:
and (3) evaluating the effect of the yaw karton diagnosis model according to the accuracy of the test set and the ROC curve: the accuracy of the test set is not less than 0.8, the area AUC under the ROC curve is calculated at the same time, the model classification performance of the yaw stuck diagnosis model measured by the ROC curve is reflected, if the AUC is not less than 80%, the model classification performance of the yaw stuck diagnosis model is proved to be good, and the step S8 is executed; and if the accuracy of the test set is less than 0.8 and the AUC is less than 80%, repeating the steps S1 to S6 until the model classification performance of the yaw Carton diagnosis model is good, and executing the step S8.
Further, the step S8 includes the steps of:
the method comprises the steps of carrying out preliminary yaw blocking identification on a yaw blocking diagnosis model, inputting actual wind turbine generator data to carry out yaw blocking identification and carrying out manual verification, carrying out manual correction if the accuracy of an actual identification result does not reach a preset accuracy, adding the actual wind turbine generator data into a training set, retraining the yaw blocking diagnosis model until the training set accuracy, the test set accuracy and the actual application identification accuracy of the yaw blocking diagnosis model are approximately the same or the same, issuing the yaw blocking diagnosis model and starting analysis to early warn yaw faults of the wind turbine generator.
The second purpose of the invention is realized by the following technical scheme: a wind turbine generator yaw blocking diagnosis system based on an SVM algorithm comprises:
the system comprises an original data acquisition module, a data processing module and a data processing module, wherein the original data acquisition module is used for acquiring sensor data recording fault actions of the wind turbine generator as original data;
the data preprocessing module is used for carrying out normalization preprocessing on the original data;
the characteristic extraction module is used for extracting the characteristics of the preprocessed original data;
the data labeling module is used for dividing the training data into a training set and a test set and labeling the training set according to a manual labeling and weak supervised learning method;
the yaw stuck diagnosis model training module is used for inputting the marked training set into an SVM algorithm for training to obtain a yaw stuck diagnosis model;
and the model online module is used for issuing a stable yaw stuck diagnosis model and starting analysis so as to early warn the yaw fault problem of the wind turbine generator.
Further, the sensor data are Tracelog data, and each Tracelog data is millisecond-level data in a time period of one minute before and after a fault action of the wind turbine generator;
screening is carried out on the obtained Tracelog data, and data of a time period recording the yawing and blocking actions of the wind turbine generator are extracted: comparing the change rate of the azimuth angles of the engine room, when the wind turbine generator is in normal yaw, the change rate of the azimuth angles of the engine room is constant, namely the change rates of the azimuth angles of the engine room at adjacent moments are equal, and the change rates of the azimuth angles of the engine room at adjacent moments are unequal when the wind turbine generator is in abnormal yaw, and at the moment, the wind turbine generator is regarded as yaw blockage.
Further, the feature extraction module comprises the steps of:
and extracting the type of azimuth angle variation of the yaw opportunity cabin, the mean value of the azimuth angle variation of the yaw opportunity cabin, the standard deviation of the azimuth angle variation of the yaw opportunity cabin, the maximum value of the azimuth angle variation of the yaw opportunity cabin, the quartile of the azimuth angle variation of the cabin, the yaw direction, the average wind speed value and the standard wind direction difference from the preprocessed original data as characteristic values.
Compared with the prior art, the invention has the following advantages and beneficial effects:
1. is scientific and effective. Compared with the traditional data analysis and detection method, the method avoids the complexity and subjectivity of determining the relevant threshold value by depending on a large amount of manual observation;
2. the accuracy is high. When the abnormal value or the unobserved information is met based on the data analysis method, misjudgment is easy to occur, and the SVM algorithm can avoid the defect by learning the distribution and the rule of the data; manually comparing data based on the algorithm diagnosis result, and issuing an early warning and checking list to the wind power plant for actual verification, wherein both the early warning and checking list and the early warning and checking list show that the SVM algorithm passes the test, and the yaw stuck diagnosis effect is good;
3. the cost is reduced, and the economic efficiency and the beneficial effect are increased. The wind turbine generator yaw system and even the transmission chain system are early warned, preventive maintenance is achieved, healthy and efficient operation of the wind turbine generator is guaranteed, the fault rate is reduced, cost is saved, and the power generation performance of a wind power plant is improved;
4. the digital operation and maintenance of the wind turbine generator is promoted. The remote monitoring is promoted to replace the conventional manual inspection, and the digital operation and maintenance era based on unattended operation, unattended operation and maintenance and predictive operation and maintenance is made a contribution.
Drawings
FIG. 1 is a flow chart of a wind turbine yaw stuck diagnostic method.
FIG. 2 is a diagram of the accuracy results of the SVM algorithm of the present invention in the training set and the test set.
FIG. 3 is a graph of ROC according to the present invention.
Detailed Description
The present invention will be further described with reference to the following specific examples.
Example 1
Referring to fig. 1, the method for diagnosing yaw seizure of a wind turbine generator based on an SVM algorithm provided in this embodiment includes the following steps:
s1, acquiring sensor data recording fault actions of a wind turbine generator as original data;
the sensor data are Tracelog data, each Tracelog data is millisecond-level data in a time period of one minute before and after a fault action of the wind turbine generator, and the data volume of each Tracelog file is 12000; for the obtained Tracelog file data at each time, two possibilities are provided, wherein firstly, no yawing action occurs in the file, and the data is generated by starting from fault actions of other components; the other type is a Tracelog file with yaw action, which may be triggered by fault action of other components or recorded by the departure of yaw fault action; therefore, whether yaw blocking occurs during the yaw action or not needs to extract time period data of the yaw action, the yaw blocking action is mainly identified by comparing the change rate of the azimuth angle of the engine room, when the wind turbine generator is in the yaw action, the change rate of the azimuth angle of the engine room in normal yaw is constant, namely the change rates of the azimuth angles of the engine rooms at adjacent moments are equal, and the change rates of the azimuth angles of the engine rooms at adjacent moments are unequal during abnormal yaw, and the engine rooms are regarded as yaw blocking.
S2, carrying out normalization pretreatment on the original data, comprising the following steps:
the method comprises the following steps that during yaw, the influence of vibration from other large components of the wind turbine generator can be caused, so that the vibration of a sensor on the wind turbine generator causes micro vibration of data of related components fixed on a cabin, the data are regarded as noise, and in order to avoid misjudgment, variables such as an azimuth angle of the cabin are filtered to eliminate the noise;
after noise filtering processing is carried out on original data, linear transformation is carried out on the filtered data through linear function normalization, the original data comprise continuous value variables and classification variables, the continuous value variables comprise the maximum value of the variation of the azimuth angle of the engine room, the mean value of the variation of the azimuth angle of the engine room, the quartile of the variation of the azimuth angle of the engine room, the average value of wind speed, the standard deviation of wind direction and the variety of the variation of the azimuth angle of the engine room, and the classification variables comprise the yaw direction, the electromagnetic band-type brake and the grid-connected mode; and mapping the continuous value variables to the range of [0,1] to realize the equal ratio scaling of each continuous value variable, wherein the normalization formula is as follows:
wherein X is the value of each continuous variable, X min Is the minimum of the values of the corresponding continuous-valued variables, X max Is the maximum value of the numerical values of the corresponding continuous value variables;
meanwhile, the yaw direction is coded, the clockwise yaw CW is recorded as 0, the counterclockwise yaw CCW is recorded as 1, and the non-yaw state is recorded as 2.
S3, extracting the characteristics of the preprocessed original data to obtain training data, and the method comprises the following steps:
and extracting the type of azimuth angle variation of the yaw opportunity cabin, the mean value of the azimuth angle variation of the yaw opportunity cabin, the standard deviation of the azimuth angle variation of the yaw opportunity cabin, the maximum value of the azimuth angle variation of the yaw opportunity cabin, the quartile of the azimuth angle variation of the cabin, the yaw direction, the average wind speed value and the standard wind direction difference from the preprocessed original data as characteristic values.
S4, dividing the training data into a training set and a test set, and labeling the training set according to a manual labeling and weak supervised learning method, wherein the method comprises the following steps:
dividing training data into a training set and a test set according to a proportion of 7; the sample case label of yaw stuck in the training set is set to 1, and the sample label of normal yaw is set to 0. In addition, if the problem of unbalanced samples between the yaw stuck sample size and the normal yaw sample size in the training set is found, the SMOTE algorithm is adopted to expand the yaw stuck sample data, so that the normal samples and the fault samples are the same in number.
S5, inputting the marked training set into an SVM algorithm for training to obtain a yaw stuck diagnosis model, wherein the SVM algorithm is excellent in performance on the training set as shown in a figure 2;
s6, performing cross validation test on the yaw stuck diagnosis model according to the test set and the ROC curve, and comprising the following steps:
referring to fig. 3, the classification capability of the yaw cartton diagnosis model is evaluated by an area AUC value calculated by an ROC curve, where an abscissa of the ROC curve is a false positive rate FPR and an ordinate is a true positive rate TPR; the formula for FPR and TPR is as follows:
the method comprises the steps that P is the number of yaw stuck data, N is the number of normal yaw data, TP is the number of yaw stuck data predicted by a yaw stuck diagnostic model in the P yaw stuck data, and FP is the number of normal yaw predicted by the yaw stuck diagnostic model in the N normal yaw data;
s7, evaluating the effect of the yaw stuck diagnosis model according to the accuracy of the test set and the ROC curve, and executing the step S8 if the evaluated effect of the yaw stuck diagnosis model meets the preset requirement; if the effect of the estimated yaw stuck diagnostic model does not meet the preset requirement, repeating the steps S1 to S6 until the effect of the estimated yaw stuck diagnostic model meets the preset requirement, and executing a step S8, wherein the step S8 comprises the following steps of:
and (3) evaluating the effect of the yaw karton diagnosis model according to the accuracy of the test set and the ROC curve: the accuracy of the test set is not less than 0.8, the area AUC under the ROC curve is calculated at the same time, the model classification performance of the yaw stuck diagnostic model measured by the ROC curve is reflected, if the AUC is not less than 80%, the model classification performance of the yaw stuck diagnostic model is good, and the step S8 is executed; and if the accuracy of the test set is less than 0.8 and the AUC is less than 80%, repeating the steps S1 to S6 until the model classification performance of the yaw Carton diagnosis model is good, and executing the step S8. By calculating the area under the ROC curve, AUC =0.973, and reflecting the model performance measured by the ROC curve, the classification performance of the yaw karton diagnosis model is proved to be better.
S8, manually checking the accuracy of the actual recognition result of the yaw stuck diagnosis model, if the actual recognition result is unqualified, adding a new data sample into a training set, and repeating the steps S5 to S7 until the actual recognition result is qualified; if the yaw blocking diagnosis model is verified to be qualified, the yaw blocking diagnosis model is issued and analysis is started to early warn yaw faults of the wind turbine generator, and the method comprises the following steps:
when the accuracy of a model training set and the accuracy of a test set are both larger than 0.8 and the difference between the accuracy of the model training set and the accuracy of the test set is not large, performing initial yaw stuck identification on a yaw stuck diagnosis model, inputting actual wind turbine generator data to perform yaw stuck identification and performing manual verification, performing manual correction if the accuracy of an actual identification result does not reach 80%, adding the actual wind turbine generator data into a training set, retraining the yaw stuck diagnosis model, and issuing the yaw stuck diagnosis model and starting analysis until the accuracy of the training set, the accuracy of the test set and the accuracy of actual application identification of the yaw stuck diagnosis model are approximately the same or the same so as to early warn the problem of yaw faults of the wind turbine generator.
Taking the Tracelog file data of a certain wind field unit in nearly one month as practical data, identifying the yaw seizure problem of the actual unit by using an SVM model, firstly acquiring a Tracelog file path corresponding to each wind turbine unit of the wind field in a database, and simultaneously setting an analysis period for a program, wherein the data of 6 months in 2022 is taken as the analysis period. In order to better illustrate the accuracy and timeliness of the SVM algorithm, the present embodiment performs yaw stuck fault detection based on a data analysis method, and compares the prediction accuracy of the yaw stuck fault detection and the prediction accuracy of the yaw stuck fault detection, and the result is shown in table 1.
TABLE 1
It is known from table 1 that the 006#, 013#, 022# units have high-frequency yaw blocking, so that the three units are checked by manual data to determine whether misjudgment occurs. The verification result shows that most of yaw stuck based on SVM model detection is correct judgment, and the data analysis method is used for diagnosing that the yaw stuck is misjudged probably because the observed data characteristics are limited.
And then, an early warning work order is issued to the site, wind field operation and maintenance personnel check the states of all parts of the yaw system, the site feedback finds that the 013# unit yaw brake is seriously worn, errors are set in the 006#, 022# unit yaw brake pressure parameters, the operation and maintenance personnel are subjected to targeted processing, the parts are prevented from being damaged more seriously, the failure rate is reduced, the promised pressure is reduced, and the loss of the generated energy caused by the failure shutdown of the unit is effectively avoided. In conclusion, the reliability of identifying the yaw seizure problem based on the SVM algorithm is high.
Example 2
The embodiment discloses a wind turbine generator yaw stuck diagnosis system based on an SVM algorithm, which comprises:
the raw data acquisition module is used for acquiring sensor data recording fault actions of the wind turbine generator and comprises the following steps of: the sensor data are Tracelog data, and each Tracelog data is millisecond-level data in a time period of one minute before and after a fault action of the wind turbine generator;
screening is carried out on the obtained Tracelog data, and data recording time periods of yawing and stopping actions of the wind turbine generator are extracted: comparing the change rate of the azimuth angles of the engine room, when the wind turbine generator is in normal yaw, the change rate of the azimuth angles of the engine room is constant, namely the change rates of the azimuth angles of the engine room at adjacent moments are equal, and the change rates of the azimuth angles of the engine room at adjacent moments are unequal when the wind turbine generator is in abnormal yaw, and at the moment, the wind turbine generator is regarded as yaw blockage.
The data preprocessing module is used for carrying out normalization preprocessing on the original data and comprises the following steps:
linear transformation is performed on the original data by linear function normalization:
the original data comprise continuous value variables and classification variables, the continuous value variables comprise the maximum value of the variation of the azimuth angle of the engine room, the mean value of the variation of the azimuth angle of the engine room, the quartile of the variation of the azimuth angle of the engine room, the mean value of the wind speed, the standard deviation of the wind direction and the variety of the variation of the azimuth angle of the engine room, and the classification variables comprise the yaw direction, the electromagnetic band-type brake and the grid-connected mode;
and mapping the continuous value variables to the range of [0,1] to realize the equal ratio scaling of each continuous value variable, wherein the normalization formula is as follows:
wherein X is the value of each continuous value variable, X min Is the minimum value of the numerical values of the corresponding continuous value variables，X max Is the maximum value of the values of the corresponding continuous-value variables.
The feature extraction module is used for extracting features of the preprocessed original data, and comprises the following steps:
extracting the type of azimuth angle variation of the yaw opportunity cabin, the mean value of the azimuth angle variation of the yaw opportunity cabin, the standard deviation of the azimuth angle variation of the yaw opportunity cabin, the maximum value of the azimuth angle variation of the yaw opportunity cabin, the quartile of the azimuth angle variation of the cabin, the yaw direction, the average wind speed and the standard wind direction difference from the preprocessed original data as characteristic values;
the data labeling module is used for dividing the training data into a training set and a test set and labeling the training set according to a manual labeling and weak supervised learning method;
the yaw stuck diagnosis model training module is used for inputting the marked training set into an SVM algorithm for training to obtain a yaw stuck diagnosis model;
the model online module is used for issuing a stable yaw stuck diagnosis model and starting analysis so as to early warn yaw faults of the wind turbine generator; when the yaw stuck diagnosis model reaches the preset accuracy, the actual unit data are verified, and when the accuracy of the actual unit data is approximately the same as that of the training set, the yaw stuck diagnosis model is considered to be stable.
The above-mentioned embodiments are merely preferred embodiments of the present invention, and the scope of the present invention is not limited thereto, so that the changes in the shape and principle of the present invention should be covered within the protection scope of the present invention.
Claims (10)
1. A wind turbine generator yaw stuck diagnosis method based on an SVM algorithm is characterized by comprising the following steps of:
s1, acquiring sensor data recording fault actions of a wind turbine generator as original data;
s2, carrying out normalization pretreatment on the original data;
s3, extracting the characteristics of the preprocessed original data to obtain model data;
s4, performing up-sampling processing on the model data, dividing the model data into a training set and a test set, and labeling the training set according to a manual labeling and weak supervision learning method;
s5, inputting the marked training set into an SVM algorithm for training to obtain a yaw stuck diagnosis model;
s6, verifying and calculating the accuracy of the test set, and constructing an ROC curve according to the normal yaw and yaw stuck data of the wind turbine generator;
s7, evaluating the effect of the yaw stuck diagnosis model according to the accuracy of the test set and the ROC curve, and executing the step S8 if the evaluated effect of the yaw stuck diagnosis model meets the preset requirement; if the effect of the estimated yaw stuck-in diagnosis model does not meet the preset requirement, repeating the steps S1 to S6 until the effect of the estimated yaw stuck-in diagnosis model meets the preset requirement, and executing the step S8;
s8, manually checking the accuracy of the actual recognition result of the yaw stuck diagnosis model, if the actual recognition result is unqualified, adding a new data sample into a training set, and repeating the steps from S5 to S7 until the actual recognition result is qualified; and if the yaw blockage diagnosis model is qualified, issuing the yaw blockage diagnosis model and starting analysis so as to early warn the yaw fault problem of the wind turbine generator.
2. The wind turbine generator yaw stuck diagnosis method based on the SVM algorithm as recited in claim 1, wherein the step S1 comprises the steps of:
the sensor data are Tracelog data, and each Tracelog data is millisecond-level data in a time period of one minute before and after a fault action of the wind turbine generator;
screening is carried out on the obtained Tracelog data, and data of a time period recording the yawing and blocking actions of the wind turbine generator are extracted: comparing the change rate of the azimuth angles of the engine rooms, when the wind turbine generator is in normal yaw, the change rate of the azimuth angles of the engine rooms is constant, namely the change rates of the azimuth angles of the engine rooms at adjacent moments are equal, and the change rates of the azimuth angles of the engine rooms at adjacent moments are unequal when the wind turbine generator is in abnormal yaw, and at the moment, the wind turbine generator is regarded as yaw blocking.
3. The SVM algorithm-based wind turbine generator yaw stuck diagnosis method as set forth in claim 1, wherein the step S2 comprises the steps of:
linear transformation is performed on the original data by linear function normalization:
the original data comprises continuous value variables and classification variables, the continuous value variables comprise the maximum value of the variation of the azimuth angle of the engine room, the mean value of the variation of the azimuth angle of the engine room, the quartile of the variation of the azimuth angle of the engine room, the mean value of wind speed, the standard deviation of wind direction and the variety of the variation of the azimuth angle of the engine room, and the classification variables comprise the yaw direction, the electromagnetic band-type brake and the grid-connected mode;
mapping the continuous value variable to the range of [0,1] to realize the equal ratio scaling of each continuous value variable;
meanwhile, the yaw direction is coded, the clockwise yaw CW is recorded as 0, the counterclockwise yaw CCW is recorded as 1, and the non-yaw state is recorded as 2.
4. The method for diagnosing the yaw seizure of the wind turbine generator set based on the SVM algorithm as recited in claim 1, wherein the step S3 comprises the steps of:
and extracting the type of azimuth angle variation of the yaw opportunity cabin, the mean value of the azimuth angle variation of the yaw opportunity cabin, the standard deviation of the azimuth angle variation of the yaw opportunity cabin, the maximum value of the azimuth angle variation of the yaw opportunity cabin, the quartile of the azimuth angle variation of the cabin, the yaw direction, the average wind speed value and the standard wind direction difference from the preprocessed original data as characteristic values.
5. The method for diagnosing the yaw stuck of the wind turbine generator based on the SVM algorithm as recited in claim 1, wherein in the step S6, the step of constructing the ROC curve according to the normal yaw and yaw stuck data of the wind turbine generator comprises the following steps:
evaluating the classification capability of the yaw Kadun diagnosis model through an area AUC value calculated by an ROC curve, wherein the abscissa of the ROC curve is a false positive rate FPR, and the ordinate is a true positive rate TPR; the formula for FPR and TPR is as follows:
wherein, P is the number of the yaw stuck data, N is the number of the normal yaw data, TP is the number of the yaw stuck data predicted by the yaw stuck diagnostic model in the P yaw stuck data, and FP is the number of the normal yaw predicted by the yaw stuck diagnostic model in the N normal yaw data.
6. The SVM algorithm-based wind turbine generator yaw stuck diagnosis method as set forth in claim 1, wherein the step S7 comprises the steps of:
and (3) evaluating the effect of the yaw stuck diagnosis model according to the accuracy of the test set and the ROC curve: the accuracy of the test set is not less than 0.8, the area AUC under the ROC curve is calculated at the same time, the model classification performance of the yaw stuck diagnostic model measured by the ROC curve is reflected, if the AUC is not less than 80%, the model classification performance of the yaw stuck diagnostic model is good, and the step S8 is executed; and if the accuracy of the test set is less than 0.8 and the AUC is less than 80%, repeating the steps S1 to S6 until the model classification performance of the yaw Carton diagnosis model is good, and executing the step S8.
7. The method for diagnosing the yaw seizure of the wind turbine generator set based on the SVM algorithm as recited in claim 1, wherein the step S8 comprises the steps of:
the method comprises the steps of carrying out preliminary yaw blocking identification on a yaw blocking diagnosis model, inputting actual wind turbine generator data to carry out yaw blocking identification and carrying out manual verification, carrying out manual correction if the accuracy of an actual identification result does not reach a preset accuracy, adding the actual wind turbine generator data into a training set, retraining the yaw blocking diagnosis model until the training set accuracy, the test set accuracy and the actual application identification accuracy of the yaw blocking diagnosis model are approximately the same or the same, issuing the yaw blocking diagnosis model and starting analysis to early warn yaw faults of the wind turbine generator.
8. A wind turbine generator yaw blocking diagnosis system based on an SVM algorithm is characterized by comprising:
the system comprises an original data acquisition module, a data processing module and a data processing module, wherein the original data acquisition module is used for acquiring sensor data recording fault actions of the wind turbine generator as original data;
the data preprocessing module is used for carrying out normalization preprocessing on the original data;
the feature extraction module is used for extracting features of the preprocessed original data;
the data labeling module is used for dividing the training data into a training set and a test set and labeling the training set according to a manual labeling and weak supervised learning method;
the yaw stuck diagnosis model training module is used for inputting the marked training set into an SVM algorithm for training to obtain a yaw stuck diagnosis model;
and the model online module is used for issuing a stable yaw stuck diagnosis model and starting analysis so as to early warn the yaw fault problem of the wind turbine generator.
9. The SVM algorithm-based wind turbine generator yaw stuck diagnosis system as claimed in claim 8, wherein the sensor data is Tracelog data, each Tracelog data being millisecond-level data within a time period of one minute before and after a fault action of the wind turbine generator;
screening is carried out on the obtained Tracelog data, and data recording time periods of yawing and stopping actions of the wind turbine generator are extracted: comparing the change rate of the azimuth angles of the engine rooms, when the wind turbine generator is in normal yaw, the change rate of the azimuth angles of the engine rooms is constant, namely the change rates of the azimuth angles of the engine rooms at adjacent moments are equal, and the change rates of the azimuth angles of the engine rooms at adjacent moments are unequal when the wind turbine generator is in abnormal yaw, and at the moment, the wind turbine generator is regarded as yaw blocking.
10. The SVM algorithm-based wind turbine generator yaw-stuck diagnostic system of claim 8, wherein the feature extraction module comprises the steps of:
and extracting the type of azimuth angle variation of the yaw opportunity cabin, the mean value of the azimuth angle variation of the yaw opportunity cabin, the standard deviation of the azimuth angle variation of the yaw opportunity cabin, the maximum value of the azimuth angle variation of the yaw opportunity cabin, the quartile of the azimuth angle variation of the cabin, the yaw direction, the average wind speed value and the standard wind direction difference from the preprocessed original data as characteristic values.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211197505.2A CN115563501A (en) | 2022-09-29 | 2022-09-29 | Wind turbine generator yaw stuck diagnosis method and system based on SVM algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211197505.2A CN115563501A (en) | 2022-09-29 | 2022-09-29 | Wind turbine generator yaw stuck diagnosis method and system based on SVM algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115563501A true CN115563501A (en) | 2023-01-03 |
Family
ID=84742806
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211197505.2A Pending CN115563501A (en) | 2022-09-29 | 2022-09-29 | Wind turbine generator yaw stuck diagnosis method and system based on SVM algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115563501A (en) |
-
2022
- 2022-09-29 CN CN202211197505.2A patent/CN115563501A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CA2969414C (en) | Apparatus and method for monitoring a device having a movable part | |
| CN111539553B (en) | Wind turbine generator fault early warning method based on SVR algorithm and off-peak degree | |
| CN113705924B (en) | Intelligent diagnosis method and system for thermal control equipment | |
| CN109681391B (en) | Blade root bolt fracture fault detection method and medium | |
| Korkos et al. | Data annotation and feature extraction in fault detection in a wind turbine hydraulic pitch system | |
| CN111103137A (en) | Wind turbine gearbox fault diagnosis method based on deep neural network | |
| CN112418306A (en) | Gas turbine compressor fault early warning method based on LSTM-SVM | |
| Zemali et al. | Robust intelligent fault diagnosis strategy using Kalman observers and neuro-fuzzy systems for a wind turbine benchmark | |
| CN114004059B (en) | Health portrait method for hydroelectric generating set | |
| Wu et al. | Condition-based monitoring and novel fault detection based on incremental learning applied to rotary systems | |
| Palmé et al. | Early warning of gas turbine failure by nonlinear feature extraction using an auto-associative neural network approach | |
| Du et al. | A SOM based Anomaly detection method for wind turbines health management through SCADA data | |
| Dutta et al. | Bearing fault detection for water pumping system using artificial neural network | |
| CN114577470A (en) | Fault diagnosis method and system for fan main bearing | |
| Gao et al. | An Overview on fault diagnosis, prognosis and resilient control for wind turbine systems. Processes. 2021; 9: 300 | |
| CN115563501A (en) | Wind turbine generator yaw stuck diagnosis method and system based on SVM algorithm | |
| Liu et al. | Fault diagnosis of diesel engine information fusion based on adaptive dynamic weighted hybrid distance-taguchi method (ADWHD-T) | |
| Souza et al. | Evaluation of data based normal behavior models for fault detection in wind turbines | |
| Eustace | A real-world application of fuzzy logic and influence coefficients for gas turbine performance diagnostics | |
| Seo et al. | Gearbox Condition Monitoring and Diagnosis of Unlabeled Vibration Signals Using a Supervised Learning Classifier | |
| Wei et al. | Multi-sensor monitoring based on-line diesel engine anomaly detection with baseline deviation | |
| Byington et al. | False alarm mitigation of vibration diagnostic systems | |
| Bilendo et al. | Wind turbine anomaly detection based on SCADA data | |
| Sun et al. | Anomaly detection of marine diesel engine valve system based on Mahalanobis distance | |
| Zhang | Condition monitoring of a fan using 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 |