CN116011332A - Wind turbine generator system state monitoring method based on GAN-QP feature migration model - Google Patents
Wind turbine generator system state monitoring method based on GAN-QP feature migration model Download PDFInfo
- Publication number
- CN116011332A CN116011332A CN202211739162.8A CN202211739162A CN116011332A CN 116011332 A CN116011332 A CN 116011332A CN 202211739162 A CN202211739162 A CN 202211739162A CN 116011332 A CN116011332 A CN 116011332A
- Authority
- CN
- China
- Prior art keywords
- data
- source domain
- target
- wind turbine
- migration model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- 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
- Wind Motors (AREA)
Abstract
The invention discloses a wind turbine generator system state monitoring method based on a GAN-QP characteristic migration model, which comprises the following steps: collecting historical data sets of the running states of the two wind turbine generator systems; taking one set as a source domain set and the other set as a target domain set, respectively preprocessing data sets of the two sets to obtain a source domain training set and a target domain training set; training a feature migration model based on the target domain training set and the source domain training set, wherein the feature migration model can transform the feature distribution of the target domain unit data to the feature distribution of the source domain unit data after training; training a self-encoder based on the source domain training set; preprocessing an online data set of a target domain unit to obtain a target domain test set, and transforming the target domain test set into a source domain space based on a trained feature migration model to obtain a source domain test set; and inputting the source domain test set into a trained self-encoder, calculating the mean square error of the source domain test set and the source domain training set, and judging whether the system operates normally or not based on the mean square error.
Description
Technical Field
The invention belongs to the technical field of wind turbine generator state monitoring, and particularly relates to a wind turbine generator state monitoring method based on a GAN-QP characteristic migration model.
Background
Along with the development of 'carbon peak, carbon neutralization', clean renewable energy represented by wind energy and solar energy is brought into new development opportunities. In particular, the wind power generation technology is continuously developed and perfected, and the newly-increased installed capacity of the wind turbine generator is steadily improved. Accordingly, in order to ensure the safe and reliable operation of the wind turbine in complex and changeable environments, developing intelligent state monitoring technology research on the wind turbine has become a hotspot problem in the wind power industry. The development of artificial intelligence technology is benefited, and the intelligent state monitoring method of the wind turbine generator based on data acquisition and supervisory control (Supervisory control and data acquisition, SCADA) data drive gradually replaces traditional state monitoring methods such as vibration monitoring and oil monitoring, and becomes a mainstream method. The method generally adopts the combination of normal behavior modeling and residual analysis to monitor the state of the wind turbine, but the training of the model needs enough training data, and the training data and the testing data need to obey independent identical distribution conditions. Therefore, the new installation unit or the unit with the lost data record cannot establish an effective model due to insufficient training data, and the method has certain limitations.
The migration learning provides a new idea for establishing a wind turbine generator system state monitoring model under the condition of insufficient data, and can extract the migratable characteristics or knowledge structures from the source fields with sufficient related data and rich field knowledge to improve the performance of the model in different but related target fields, so that the requirements that training data must be sufficient and the training data and the testing data are subjected to independent and same distribution are relaxed to a certain extent.
Disclosure of Invention
The invention aims to provide a wind turbine generator system state monitoring method based on a feature migration model, the feature migration method based on the feature migration model can still realize knowledge multiplexing under the condition of large data distribution difference between a source domain and a target domain, and further the feature migration method is combined with countermeasure learning, so that the distribution adaptation between the source domain and the target domain data can be effectively promoted, the requirements that training data must be sufficient and test data are subjected to independent same distribution are relaxed to a certain extent, and a new idea is provided for newly installing a turbine generator system or a turbine generator system with data record loss, and building the wind turbine generator system state monitoring model under the condition of insufficient data.
In order to achieve the above purpose, the invention adopts the following technical scheme:
a wind turbine generator system state monitoring method based on a GAN-QP feature migration model comprises the following steps:
s1, collecting a generator system operation state historical data set of two wind turbines with the same type but different geographic positions and rated powers;
s2, taking one wind turbine generator set as a source domain unit and the other wind turbine generator set as a target domain unit, and respectively preprocessing a historical data set of the running states of the generator systems of the two wind turbine generator sets to obtain a source domain training set and a target domain training set;
s3, training the feature migration model based on the target domain training set and the source domain training set to obtain a trained feature migration model, wherein the trained feature migration model can transform the feature distribution of the target domain unit operation data to the feature distribution of the source domain unit operation data;
s4, training the self-encoder based on the source domain training set to learn the characteristic distribution rule of the data in the normal running state of the source domain unit so as to obtain the trained self-encoder;
s5, preprocessing an online data set of the running state of the generator system of the target domain unit to obtain a target domain test set, and transforming the target domain test set into a source domain space based on the trained feature migration model to obtain a source domain test set;
s6, taking the source domain test set obtained in the step S5 as input data of the self-encoder after training, calculating the mean square error of the source domain test set and the source domain training set, and judging whether the running state of the generator system is normal or not based on the mean square error.
In a preferred scheme, in step S2, the preprocessing of the historical data sets of the running states of the generator systems of the source domain unit and the target domain unit includes the following steps:
s2.1, deleting abnormal data in the historical data set; :
and S2.2, carrying out normalization processing on the historical data set processed in the step S2.1.
In the preferred scheme, in step S2.1, the abnormal data includes data with missing values, shutdown data, data exceeding a wind speed range of normal operation of the wind turbine, and outlier data.
Preferably, the determining of the outlier data includes the steps of:
A. selecting a certain data point in the historical data set;
B. selecting a plurality of target data points which are closest to the data points selected in the step A in the historical data set;
C. calculating an average chain distance for the selected data point based on the selected data point in step a and the plurality of target data points;
D. calculating the connection anomaly factor-based distance of the selected data point based on the average chain distance of the selected data point and the average chain distances corresponding to the plurality of target data points;
E. based on the connection anomaly factor-based distance of the selected data point, it is determined whether the selected data point is outlier data.
Preferably, the step D includes the steps of:
calculating average chain distances corresponding to the target data points respectively;
calculating an average of the average chain distances of the plurality of target data points;
based on the average chain distance of the selected data points and the average value of the average chain distances of the plurality of target data points, the connection anomaly factor-based distance of the selected data points is calculated.
Preferably, the average chain distance of the selected data points
The calculation formula is as follows:
where k is the total number of target data points, e i Representing the distance of the ith target data point to the selected data point.
Preferably, the selected data points are based on a link abnormality factor distance COF k The calculation formula of (p) is:
wherein N is k (p) represents a plurality of target data point sets,
representing the average chain distance of the o-th target data point.
Preferably, the step S3 includes the following steps:
s3.1, dividing the target domain training set into a plurality of batches;
s3.2, selecting a certain batch of target domain training sets to input into a generator network row for preprocessing so as to obtain a generated data set;
s3.3, inputting the generated data set and the source domain training set into a discriminator network to output discrimination results of the generated data set and the source domain training set;
s3.4, calculating loss function values of the generator network and the discriminator network based on the discrimination result, and updating model parameters of the generator network and the discriminator network based on the loss function values;
s3.5, circularly executing the steps S3.2-S3.4 to the preset times, so that the characteristic distribution of the generated data set of the generator network approximates to the characteristic distribution of the source domain training set, and taking the finally obtained generator network as a trained characteristic migration model.
Preferably, the loss function L used by the arbiter network D The method comprises the following steps:
wherein D (·) represents the arbiter output, x g Representing the generation of dataset data, x r Represents source domain training set data, lambda is a balance coefficient, d (x r ,x g ) Represents x r And x g Euclidean distance between, p r For source domain training set data distribution, p g In order to generate the data distribution of the data set,
an expected value is calculated for each sample data in the source domain training set data distribution and the generated data set data distribution.
Preferably, the generator network employs a loss function L G The method comprises the following steps:
the beneficial effects of the invention are as follows:
aiming at the problem that the running data of a newly installed or lost wind turbine is insufficient and an effective traditional machine learning monitoring model cannot be trained, the invention provides a wind turbine state monitoring method based on a GAN-QP characteristic migration model.
The invention provides a new idea for establishing an effective state monitoring model of the mechanical equipment under the condition of data missing.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a wind turbine generator system state monitoring method based on a GAN-QP feature migration model.
FIG. 2 is a schematic diagram of source domain and target domain set data distribution adaptation based on a feature migration model.
FIG. 3 is a plot of wind speed versus active power for a source domain unit and a target domain unit generator system with historical data before and after pretreatment, respectively.
FIG. 4 is a plot of wind speed versus generator speed versus active power for the source domain unit and the target domain unit generator system state history data before and after a distribution adaptation.
FIG. 5 is a graph of the condition monitoring of a generator of a target field unit under no migration conditions and under migration conditions using the method of the present invention.
Detailed Description
The following specific examples are presented to illustrate the present invention, and those skilled in the art will readily appreciate the additional advantages and capabilities of the present invention as disclosed herein. The invention may be practiced or carried out in other embodiments that depart from the specific details, and the details of the present description may be modified or varied from the spirit and scope of the present invention. It should be noted that the following embodiments and features in the embodiments may be combined with each other without conflict.
Referring to fig. 1, a wind turbine generator system state monitoring method based on a GAN-QP feature migration model includes the following steps:
s1, collecting a generator system operation state historical data set of two wind turbines with the same type but different geographic positions and rated powers;
s2, taking one wind turbine generator set as a source domain unit and the other wind turbine generator set as a target domain unit, and respectively preprocessing a historical data set of the running states of the generator systems of the two wind turbine generator sets to obtain a source domain training set and a target domain training set;
s3, training the feature migration model based on the target domain training set and the source domain training set to obtain a trained feature migration model, wherein the trained feature migration model can transform the feature distribution of the target domain unit operation data to the feature distribution of the source domain unit operation data;
s4, training the self-encoder based on the source domain training set to learn the characteristic distribution rule of the data in the normal running state of the source domain unit so as to obtain the trained self-encoder;
s5, preprocessing an online data set of the running state of the generator system of the target domain unit to obtain a target domain test set, and transforming the target domain test set into a source domain space based on the trained feature migration model to obtain a source domain test set;
s6, taking the source domain test set obtained in the step S5 as input data of the self-encoder after training, calculating the mean square error of the source domain test set and the source domain training set, and judging whether the running state of the generator system is normal or not based on the mean square error.
Further, in step S1, data relating to the operating state of the generator system is extracted as shown in the following table:
further, in step S2, the preprocessing of the historical data sets of the generator system running states of the source domain unit and the target domain unit includes the following steps:
s2.1, deleting abnormal data in the historical data set;
s2.2, carrying out normalization processing on the historical data set processed in the step S2.1, and eliminating the influence of dimension so as to reduce the model training difficulty and improve the model accuracy.
In step S2.1, the abnormal data includes data with a missing value, shutdown data, data exceeding a normal operation wind speed interval of the wind turbine, and outlier data.
Specifically:
the shutdown data are data with active power of 0; the data exceeding the normal operation wind speed interval of the wind turbine generator are data with wind speed smaller than cut-in wind speed (3 m/s) and data with wind speed larger than cut-out wind speed (25 m/s); the data of the limited power operation are data of which the power is constantly distributed below the rated power and does not change along with the wind speed, and the data are mainly obtained through manual control and cannot correctly reflect the operation rule of the unit.
FIG. 3 (a) is a plot of unit wind speed versus active power scatter of source domain unit generator system state history data prior to preprocessing; FIG. 3 (b) is a plot of wind speed versus active power scatter of source field unit generator system state history data after preprocessing; FIG. 3 (c) is a wind speed-active power scatter plot of target domain unit generator system state history data prior to preprocessing; FIG. 3 (d) is a graph of wind speed versus active power scatter of target domain unit generator system state history data after preprocessing.
The source domain training set and the target domain training set obtained by preprocessing the data set are both normal data sets.
Further, the determining of the outlier data includes the steps of:
A. selecting a certain data point in the historical data set;
B. selecting a plurality of target data points which are closest to the data points selected in the step A in the historical data set;
C. calculating an average chain distance for the selected data point based on the selected data point in step a and the plurality of target data points;
D. calculating the connection anomaly factor-based distance of the selected data point based on the average chain distance of the selected data point and the average chain distances corresponding to the plurality of target data points;
E. based on the connection anomaly factor-based distance of the selected data point, it is determined whether the selected data point is outlier data.
Further, in step D, the following steps are included:
calculating average chain distances corresponding to the target data points respectively;
calculating an average of the average chain distances of the plurality of target data points;
based on the average chain distance of the selected data points and the average value of the average chain distances of the plurality of target data points, the connection anomaly factor-based distance of the selected data points is calculated.
Further, the average chain distance of the selected data points
The calculation formula is as follows:
where k is the total number of target data points, e i Representing the distance of the ith target data point to the selected data point.
Further, the selected data points are based on a connection anomaly factor distance COF k The calculation formula of (p) is:
wherein N is k (p) represents a plurality of target data point sets,
representing the average chain distance of the o-th target data point.
Specifically:
the value of COF is the possibility that point p is an outlier, and when the value of COF is less than 1, this point is indicated as a normal value, and when the value of COF is 1 or more, this point is indicated as an outlier.
Further, the formula for normalizing the historical data set processed in step S2.1 is as follows:
wherein x is the historical data set data processed in the step S2.1, min (x) is the minimum value of x, and max (x) is the maximum value of x.
Further, after the source domain training set and the target domain training set are obtained through data preprocessing, a step S3 is executed, and in the step S3, the method includes the following steps:
s3.1, dividing the target domain training set into a plurality of batches;
s3.2, selecting a certain batch of target domain training sets to input into a generator network row for preprocessing so as to obtain a generated data set;
s3.3, inputting the generated data set and the source domain training set into a discriminator network to output discrimination results of the generated data set and the source domain training set;
s3.4, calculating loss function values of the generator network and the discriminator network based on the discrimination result, and updating model parameters of the generator network and the discriminator network based on the loss function values;
s3.5, circularly executing the steps S3.2-S3.4 to the preset times, so that the characteristic distribution of the generated data set of the generator network approximates to the characteristic distribution of the source domain training set, and taking the finally obtained generator network as a trained characteristic migration model.
Further, a loss function L adopted by the discriminator network D The method comprises the following steps:
wherein D (·) represents the arbiter output, x g Representing the generation of dataset data, x r Represents source domain training set data, lambda is a balance coefficient, d (x r ,x g ) Represents x r And x g Euclidean distance between, p r For source domain training set data distribution, p g In order to generate the data distribution of the data set,
an expected value is calculated for each sample data in the source domain training set data distribution and the generated data set data distribution.
Further, a loss function L employed by the generator network G The method comprises the following steps:
specifically:
and carrying out distribution adaptation on the source domain and target domain unit data based on the trained feature migration model, wherein the data distribution of the wind speed-generator rotating speed-active power of the source domain unit and the target domain unit before distribution adaptation is shown in the figure 4 (a), and the data distribution of the wind speed-generator rotating speed-active power of the source domain unit and the target domain unit after distribution adaptation is shown in the figure 4 (b).
According to the method, the characteristic migration model can be trained based on the target domain training set and the source domain training set, the obtained trained characteristic migration model can transform the characteristic distribution of the target domain unit operation data into the characteristic distribution of the source domain unit operation data, the distribution adaptation between the source domain and the target domain data can be effectively promoted, the requirements that the training data must be sufficient and the test data are required to be subjected to independent identical distribution are relaxed to a certain extent, and a new thought is provided for newly installing a unit or a unit with lost data record and establishing a wind turbine unit state monitoring model under the condition of insufficient data.
More specifically:
the mean square error calculated in the step S6 is used for judging whether the running state of the generator system is normal or not.
Because the wind field is generally located in remote areas such as mountain coasts and the like, maintenance is inconvenient, a loose early warning strategy is selected to avoid frequent warning, the source field test set obtained in the step S5 is used as input data of a self-encoder after training, the mean square error of the source field test set and the source field training set is calculated, and the maximum value of the obtained mean square error is used as a health threshold value, so that early warning frequency is reduced, and the operation and maintenance efficiency of the wind field is improved.
Because the source domain training set and the target domain training set are all health data of the unit, the mean square error obtained by taking the health data as input data is all judged to be normal, and therefore, the early warning frequency can be reduced by adopting the maximum value as a monitoring threshold.
The method provided by the invention can be used for effectively monitoring whether the running state of the generator of the wind turbine generator is normal or not, has low early warning frequency and can improve the operation and maintenance efficiency of a wind power plant.
The above examples are merely illustrative of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention, and various modifications and improvements made by those skilled in the art to the technical solution of the present invention should fall within the protection scope of the present invention without departing from the design spirit of the present invention.
Claims (10)
1. A wind turbine generator system state monitoring method based on a GAN-QP feature migration model is characterized by comprising the following steps:
s1, collecting a generator system operation state historical data set of two wind turbines with the same type but different geographic positions and rated powers;
s2, taking one wind turbine generator set as a source domain unit and the other wind turbine generator set as a target domain unit, and respectively preprocessing a historical data set of the running states of the generator systems of the two wind turbine generator sets to obtain a source domain training set and a target domain training set;
s3, training the feature migration model based on the target domain training set and the source domain training set to obtain a trained feature migration model, wherein the trained feature migration model can transform the feature distribution of the target domain unit operation data to the feature distribution of the source domain unit operation data;
s4, training the self-encoder based on the source domain training set to learn the characteristic distribution rule of the data in the normal running state of the source domain unit so as to obtain the trained self-encoder;
s5, preprocessing an online data set of the running state of the generator system of the target domain unit to obtain a target domain test set, and transforming the target domain test set into a source domain space based on the trained feature migration model to obtain a source domain test set;
s6, taking the source domain test set obtained in the step S5 as input data of the self-encoder after training, calculating the mean square error of the source domain test set and the source domain training set, and judging whether the running state of the generator system is normal or not based on the mean square error.
2. The method for monitoring the state of a wind turbine generator system based on the GAN-QP feature migration model according to claim 1, wherein in step S2, the preprocessing of the historical data sets of the running states of the generator systems of the source domain unit and the target domain unit includes the following steps:
s2.1, deleting abnormal data in the historical data set; :
and S2.2, carrying out normalization processing on the historical data set processed in the step S2.1.
3. The method for monitoring the state of the wind turbine generator based on the GAN-QP feature migration model according to claim 2, wherein in step S2.1, the abnormal data includes data with a missing value, shutdown data, data exceeding a normal operation wind speed interval of the wind turbine generator, and outlier data.
4. A method for monitoring the state of a wind turbine generator based on a GAN-QP feature migration model according to claim 3, wherein the determining of the outlier data includes the steps of:
A. selecting a certain data point in the historical data set;
B. selecting a plurality of target data points which are closest to the data points selected in the step A in the historical data set;
C. calculating an average chain distance for the selected data point based on the selected data point in step a and the plurality of target data points;
D. calculating the connection anomaly factor-based distance of the selected data point based on the average chain distance of the selected data point and the average chain distances corresponding to the plurality of target data points;
E. based on the connection anomaly factor-based distance of the selected data point, it is determined whether the selected data point is outlier data.
5. The method for monitoring the state of a wind turbine generator based on the GAN-QP feature migration model according to claim 4, wherein in step D, the method comprises the steps of:
calculating average chain distances corresponding to the target data points respectively;
calculating an average of the average chain distances of the plurality of target data points;
based on the average chain distance of the selected data points and the average value of the average chain distances of the plurality of target data points, the connection anomaly factor-based distance of the selected data points is calculated.
6. The method for monitoring the state of a wind turbine generator based on a GAN-QP feature migration model of claim 5, wherein the average chain distance of selected data points
The calculation formula is as follows:
where k is the total number of target data points, e i Representing the distance of the ith target data point to the selected data point.
7. The method for monitoring the state of a wind turbine generator based on a GAN-QP feature migration model of claim 6, wherein the selected data points are based on a connection anomaly factor distance COF k Calculation formula of (p)The formula is:
wherein N is k (p) represents a plurality of target data point sets,
representing the average chain distance of the o-th target data point.
8. The method for monitoring the state of a wind turbine generator based on the GAN-QP feature migration model according to claim 1, wherein in step S3, the method comprises the following steps:
s3.1, dividing the target domain training set into a plurality of batches;
s3.2, selecting a certain batch of target domain training sets to input into a generator network row for preprocessing so as to obtain a generated data set;
s3.3, inputting the generated data set and the source domain training set into a discriminator network to output discrimination results of the generated data set and the source domain training set;
s3.4, calculating loss function values of the generator network and the discriminator network based on the discrimination result, and updating model parameters of the generator network and the discriminator network based on the loss function values;
s3.5, circularly executing the steps S3.2-S3.4 to the preset times, so that the characteristic distribution of the generated data set of the generator network approximates to the characteristic distribution of the source domain training set, and taking the finally obtained generator network as a trained characteristic migration model.
9. The method for monitoring the state of a wind turbine generator based on a GAN-QP feature migration model according to claim 8, wherein the loss function L is used by the arbiter network D The method comprises the following steps:
wherein D (·) represents the arbiter output, x g Representing the generation of dataset data, x r Represents source domain training set data, lambda is a balance coefficient, d (x r ,x g ) Represents x r And x g Euclidean distance between, p r For source domain training set data distribution, p g In order to generate the data distribution of the data set,
an expected value is calculated for each sample data in the source domain training set data distribution and the generated data set data distribution.
10. The method for monitoring the state of a wind turbine generator based on a GAN-QP feature migration model according to claim 9, wherein the generator network uses a loss function L G The method comprises the following steps:
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211739162.8A CN116011332A (en) | 2022-12-31 | 2022-12-31 | Wind turbine generator system state monitoring method based on GAN-QP feature migration model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211739162.8A CN116011332A (en) | 2022-12-31 | 2022-12-31 | Wind turbine generator system state monitoring method based on GAN-QP feature migration model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116011332A true CN116011332A (en) | 2023-04-25 |
Family
ID=86022692
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211739162.8A Pending CN116011332A (en) | 2022-12-31 | 2022-12-31 | Wind turbine generator system state monitoring method based on GAN-QP feature migration model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116011332A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118013443A (en) * | 2024-04-08 | 2024-05-10 | 华侨大学 | Online real-time vacuum dry pump abnormality detection method based on generation model algorithm |
-
2022
- 2022-12-31 CN CN202211739162.8A patent/CN116011332A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118013443A (en) * | 2024-04-08 | 2024-05-10 | 华侨大学 | Online real-time vacuum dry pump abnormality detection method based on generation model algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Park et al. | Development of a novel power curve monitoring method for wind turbines and its field tests | |
| Qian et al. | Data-driven condition monitoring approaches to improving power output of wind turbines | |
| CN110259646B (en) | Wind generating set component state early warning method based on historical data | |
| CN108072524B (en) | Wind turbine generator gearbox bearing fault early warning method | |
| Hu et al. | Using domain knowledge features for wind turbine diagnostics | |
| CN113591359A (en) | Cut-in/cut-out wind speed adjusting and optimizing method, system and equipment medium of wind turbine generator | |
| CN103925155A (en) | Self-adaptive detection method for abnormal wind turbine output power | |
| CN110544184A (en) | fan performance analysis method based on operation parameters | |
| Du et al. | A SCADA data based anomaly detection method for wind turbines | |
| Sun et al. | Abnormal detection of wind turbine operating conditions based on state curves | |
| CN109800931A (en) | Wind power plant generated energy loss measurement method and system based on blower SCADA data | |
| CN117149498B (en) | Power plant fault diagnosis method and system | |
| CN116011332A (en) | Wind turbine generator system state monitoring method based on GAN-QP feature migration model | |
| CN116609055A (en) | Method for diagnosing wind power gear box fault by using graph convolution neural network | |
| CN113761692B (en) | Method for identifying running states of multiple wind turbine generators based on migration component analysis | |
| Rashid et al. | Anomaly Detection of Wind Turbine Gearbox based on SCADA Temperature Data using Machine Learning | |
| CN114607571A (en) | Offshore wind power gear box fault identification method and system based on lubricating system monitoring | |
| CN110334951B (en) | Intelligent evaluation method and system for high-temperature capacity reduction state of wind turbine generator | |
| Ristić et al. | Digital technologies emergence in the contemporary hydropower plants operation | |
| CN111794921B (en) | Onshore wind turbine blade icing diagnosis method based on migration component analysis | |
| CN111178601B (en) | Wind turbine generator power prediction method based on meteorological data post-processing | |
| CN113991647A (en) | Frequency response capacity planning-oriented random production simulation method for power system | |
| Ristić et al. | A short overview on Industry 4.0 in maintenance of hydropower plants | |
| Qian et al. | A weighted kNN fault detection based on multistep index and dynamic neighborhood scale under complex working conditions | |
| CN113431800A (en) | Method for distinguishing running state of fan in real time |
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 |