CN114444382A - Wind turbine generator gearbox fault diagnosis and analysis method based on machine learning algorithm - Google Patents
Wind turbine generator gearbox fault diagnosis and analysis method based on machine learning algorithm Download PDFInfo
- Publication number
- CN114444382A CN114444382A CN202111674258.6A CN202111674258A CN114444382A CN 114444382 A CN114444382 A CN 114444382A CN 202111674258 A CN202111674258 A CN 202111674258A CN 114444382 A CN114444382 A CN 114444382A
- Authority
- CN
- China
- Prior art keywords
- data
- gearbox
- wind turbine
- turbine generator
- value
- 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.)
- Granted
Links
- 238000004458 analytical method Methods 0.000 title claims abstract description 33
- 238000003745 diagnosis Methods 0.000 title claims abstract description 30
- 238000010801 machine learning Methods 0.000 title claims abstract description 19
- 238000012545 processing Methods 0.000 claims abstract description 45
- 230000002159 abnormal effect Effects 0.000 claims abstract description 33
- 238000007781 pre-processing Methods 0.000 claims abstract description 14
- 230000004913 activation Effects 0.000 claims description 18
- 238000010276 construction Methods 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 18
- 239000003921 oil Substances 0.000 claims description 16
- 238000010219 correlation analysis Methods 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 9
- 239000010687 lubricating oil Substances 0.000 claims description 8
- 238000012549 training Methods 0.000 claims description 7
- 238000011176 pooling Methods 0.000 claims description 6
- 238000012163 sequencing technique Methods 0.000 claims description 6
- 238000012795 verification Methods 0.000 claims description 6
- 230000001133 acceleration Effects 0.000 claims description 4
- 230000005856 abnormality Effects 0.000 claims description 3
- 238000012217 deletion Methods 0.000 claims description 3
- 230000037430 deletion Effects 0.000 claims description 3
- 230000006872 improvement Effects 0.000 description 11
- 230000008034 disappearance Effects 0.000 description 8
- 238000000034 method Methods 0.000 description 7
- 230000000694 effects Effects 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000012423 maintenance Methods 0.000 description 4
- 238000013507 mapping Methods 0.000 description 4
- 230000015556 catabolic process Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- 238000006731 degradation reaction Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000010248 power generation Methods 0.000 description 3
- 230000002035 prolonged effect Effects 0.000 description 3
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 239000012535 impurity Substances 0.000 description 2
- 238000009434 installation Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000013021 overheating Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000008929 regeneration Effects 0.000 description 1
- 238000011069 regeneration method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
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
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/08—Thermal analysis or thermal optimisation
-
- 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
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Educational Administration (AREA)
- Marketing (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Tourism & Hospitality (AREA)
- Development Economics (AREA)
- General Business, Economics & Management (AREA)
- Game Theory and Decision Science (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Quality & Reliability (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- Operations Research (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Wind Motors (AREA)
Abstract
The invention discloses a wind turbine generator gearbox fault diagnosis and analysis method based on a machine learning algorithm, which comprises the following steps: analyzing actual operation data of a certain wind power plant collected by an SCADA system, and collecting data every ten minutes, wherein the data types comprise wind speed, wind power, oil temperature of a gear box, wind direction, yaw angle, pitch angle, voltage, current and the like; data preprocessing: null and outlier processing; selecting the characteristic with high correlation with the high-speed bearing of the gearbox in the data through the Pearson correlation coefficient; and constructing a gearbox fault prediction model by using a ResNet network, and predicting the temperature of the high-speed bearing of the gearbox. Evaluating the degree of deviation of the gearbox from a normal state through a residual between a predicted value and an actual value of the high-speed bearing temperature of the gearbox, setting upper and lower limit values of the residual according to a 3-sigma rule to perform early warning of abnormal temperature faults of the high-speed shaft of the gearbox, and performing early warning diagnosis and analysis by workers according to fault early warning data and the combination of field conditions.
Description
Technical Field
The invention relates to the technical field of wind driven generators, in particular to a wind turbine generator gearbox fault diagnosis and analysis method based on a machine learning algorithm.
Background
In order to meet the requirement, new energy sources must be searched to replace traditional energy sources, wind energy is used as a novel clean pollution-free energy source, and the wind energy is widely applied to the field of power generation due to the characteristics of regeneration, abundant reserves and the like. The new installation of Chinese wind power generation accounts for 75% of the global new installation. Therefore, the technical key of wind power generation is to ensure the efficient and safe operation of the wind generating set, and the fault early warning and diagnosis of the gear box is the key to ensure the safe operation of the wind power generator.
The high-speed bearing of the gear box is one of main moving parts in the gear box of the wind turbine generator, and the high-speed bearing of the gear box can affect the production efficiency of equipment once a fault occurs, even can cause the shutdown and maintenance of the whole generator set, and brings unnecessary economic loss. The high-speed bearing of the wind turbine generator gearbox is in a high-speed, high-temperature and variable-load working environment for a long time, and the operation and maintenance difficulty is high. Therefore, fault diagnosis and early warning are carried out on the high-speed bearing of the transmission chain of the wind turbine generator set so as to ensure safe and stable operation of equipment, which is also the requirement and development trend of the current industrialization process. Compared with fault diagnosis, the research make internal disorder or usurp meaning of the high-speed bearing fault early warning of the wind turbine gearbox is larger.
Disclosure of Invention
The invention aims to provide a wind turbine generator gearbox fault diagnosis and analysis method based on a machine learning algorithm, so as to solve the problem that the production efficiency of equipment is influenced due to the fault of a high-speed bearing of a gearbox, and even the huge economic loss caused by shutdown and maintenance of the whole generator is caused.
In order to achieve the above purpose, the invention provides the following technical scheme:
a wind turbine generator gearbox fault diagnosis and analysis method based on a machine learning algorithm comprises the following steps:
s101 sparse data construction: according to a set first time frequency, acquiring a plurality of types of operation parameters related to a gearbox of the wind turbine generator in the first time frequency from historical data and real-time data; s102, data preprocessing: respectively carrying out null value processing and abnormal value processing on each type of operation parameters;
s103, correlation analysis: selecting an operation parameter with high correlation with a high-speed bearing of the gearbox from a plurality of types of operation parameters through a Pearson correlation coefficient;
s104, construction of a high-speed bearing temperature prediction model of the wind turbine generator gearbox: constructing a high-speed bearing temperature prediction model of the wind turbine generator gearbox by utilizing a Resnet18 network and the operation parameters with higher correlation, and outputting a predicted value;
s105, constructing an early warning rule: and constructing an early warning rule for the deviation degree of the gearbox from the normal state by utilizing a residual between a predicted value and an actual value of the high-speed bearing temperature of the gearbox, and setting upper and lower limit values of the residual according to a 3-sigma rule to early warn the fault of the high-speed shaft temperature abnormality of the gearbox.
S106, early warning diagnosis analysis: and the staff performs early warning diagnosis analysis according to the fault early warning data and the field condition.
In the technical scheme, sparse data is constructed firstly, and if real-time data is used for analysis to acquire the correlation between all data characteristics and the high-speed bearing of the gearbox of the wind turbine generator, the data volume is overlarge, and the efficiency is low. Sparse data is selected here.
In the technical scheme, correlation analysis is added, so that the operation parameters related to the core of the wind turbine gearbox can be screened out, redundant useless or low-correlation data are eliminated, operation is reduced, and early warning efficiency is improved.
According to the technical scheme, after the temperature of the high-speed bearing of the gearbox of the wind turbine generator is predicted, residual errors are compared with actual values, and multiple times of comparison are carried out, so that the early warning accuracy is higher, and the early warning caused by overhigh temperature in other special conditions is avoided.
In the technical scheme, the operation state of the wind turbine generator gearbox high-speed bearing is directly reflected through early warning of the temperature of the high-speed bearing of the wind turbine generator gearbox, for example, the rotating speed is too high, or the overload is caused, the temperature can be directly fed back, and then early warning of other related parameters can be directly performed through data, so that early warning of less data is realized.
As a further improvement of the present invention, the S101 sparse data construction specifically includes: and taking 10min as a first time frequency, acquiring all characteristics of a wind field site as a plurality of types of operation parameters, wherein all characteristics comprise wind speed, wind power, gear box oil temperature, wind direction, yaw angle, pitch angle, voltage, current, instantaneous rotating speed of a generator, power grid active power, gear box intermediate shaft driving end bearing temperature, gear box intermediate shaft non-driving end bearing temperature, motor side gear box high speed shaft bearing temperature, gear box lubricating oil pool temperature, gear box lubricating oil inlet temperature and gear box oil way filter screen front oil pressure.
In the technical scheme, 8-15min is used as a frequency, if the frequency is too low, the detection is likely to be too frequent, the data continuity is too strong, the use is affected due to the fact that the acquisition unit is overheated, and if the time is too long and is more than 15min, the data cannot be acquired in time, and the early warning is performed only when the problem occurs. Furthermore, the selected operation parameters directly influence the temperature of the rotating shaft of the motor.
As a further improvement of the present invention, after the step S101 of constructing the sparse data, before the step S102 of preprocessing the data, the method further includes processing the operation parameters, specifically: and taking the average value, the maximum value and the minimum value of each type of operation parameters in the first time frequency.
In the technical scheme, if the correlation between all data characteristics and the high-speed bearing of the gearbox of the wind turbine generator is required to be obtained and the real-time data is used for analysis, the data volume is overlarge and the efficiency is low. Sparse data with a frequency of 8-15 minutes is used here.
As a further improvement of the present invention, in the data preprocessing of step S102, the null value processing specifically includes: and sorting each type of operation parameters according to rows, directly adopting row deletion processing for continuously more than 5 null values in each row, and otherwise, replacing the null values with 0 values.
In the technical scheme, as the evacuation data is controlled more, the individual null value processing is required, the purity of the data is ensured, and the impurity data and the like are reduced.
As a further improvement of the present invention, in the step S102 data preprocessing, the abnormal value processing specifically includes: and adopting a KNN clustering algorithm, taking K as a radius, wherein the K is a natural number more than or equal to 3, selecting K operation parameters closest to the operation parameter to be predicted for class distinguishing, and if the operation parameter to be predicted does not belong to any one of a plurality of classes of operation parameters, deleting the operation parameter as an abnormal value.
In the technical scheme, the general approximate value of K is 10, and outlier noise abnormal data with low density in the original data are removed by adopting a KNN clustering algorithm. The basic principle of the algorithm is to assume that similar objects are very close to each other, the distance between the two objects is used for explaining the similarity, and the smaller the distance is, the higher the similarity is.
As a further improvement of the present invention, in the step S102 data preprocessing, the abnormal value processing specifically includes the following steps:
first step K value selection: taking historical data as sample data, splitting the sample data into training data and verification data according to a proportion, and selecting a proper K value through cross verification, wherein K is a natural number which is more than or equal to 3 and less than or equal to 30;
and a second step of data input: inputting a plurality of types of operation parameters in the historical data as tag data;
and thirdly, processing operation parameters: for each operation parameter x in each class, performing class judgment, specifically including:
3.1 distance calculation: calculating a distance data set q between the operation parameter x and other label data, wherein q is [ q ═ q [ [ q ] of any operation parameter x1,q2,...,qm]Wherein q1 represents the distance of the first label parameter from the operation parameter x, and qm represents the distance of the last label parameter from the operation parameter x;
3.2 sequencing: sequencing the obtained distance data sets q from small to large, and selecting the label data positioned at the first K;
3.3 classification: classifying the operation parameters x into A, wherein A is K label data and the operation parameter class with the most occurrence times;
and fourthly, processing abnormal data: and circularly performing the third step, taking the operation parameters which do not belong to any kind of operation parameters as abnormal data, and deleting the abnormal data.
In the technical scheme, the abnormal value processing is realized through the classification, and the abnormal value processing performed by the clustering algorithm can effectively exclude the abnormal value. In this embodiment, what kind of (4) includes input characteristics such as wind speed, wind power, gearbox oil temperature, wind direction, yaw angle, pitch angle, voltage, and current.
As a further improvement of the present invention, the step S103 of correlation analysis comprises the following steps: assuming two variables are present in the sample, then
Wherein x isiAnd yiThe values of variables x and y at time i,
and
is the average of x and y, respectively, xiGearbox high speed bearing temperature at time i, yiSome kind of operating parameter, r, at time iiIs when i isEngraving xiAnd y ofiAnd calculating the operation parameter with the correlation degree r larger than 0.6.
According to the technical scheme, the characteristics with high relevance to the high-speed bearing of the gearbox are selected through calculation of the relevance degree, the characteristics are used as input of a wind turbine generator gearbox high-speed bearing temperature prediction model, and the wind turbine generator gearbox high-speed bearing temperature can be accurately fitted.
As a further improvement of the invention, the construction of the high-speed bearing temperature prediction model of the gearbox of the S104 wind turbine generator is specifically as follows:
the first step is as follows: the operating parameters are used as input characteristic data, and the operation parameters are subjected to standardization processing, namely 0 mean value and 1 variance.
The second step is that: and constructing a Resnet18 network by using at least five convolutional layers, one maximum pooling layer and an average pooling layer, wherein in the Resnet18 network, the operation parameters after standardized processing are used as input, and an output value in a sigmod function form is obtained and used as a predicted value.
In the technical scheme, the plurality of convolution layers are arranged to form a deeper network, so that finer functions can be captured, and a better fitting effect is obtained.
As a further improvement of the present invention, the step S104 further includes adding a short connection, specifically: in the convolutional layers before the activation of the activation function ReLU, a short connection is inserted with the first and last convolutional layers as the connection.
In the technical scheme, along with the increase of the number of network layers, a plurality of problems such as information loss caused by the gradient disappearance problem exist, the gradient obtained by the activation output of each processing layer of the network is often small, so that the propagation of the activation output and the gradient is poor, and the cost function calculation process is prolonged. This situation can lead to a rapid degradation of the accuracy of the convolutional neural network after optimization. Aiming at the problems, the temperature of the high-speed bearing of the gearbox of the wind generating set in the time period is predicted through the ResNet18 network, and the phenomenon of gradient disappearance caused by too deep network layers can be effectively avoided. In general neural networks, the deeper the network layer number is, the better the training effect is, but an over-fitting phenomenon and a gradient disappearance phenomenon occur when the depth is too deep, the over-fitting phenomenon is good in the training result, the testing result is poor, however, the resnet18 network increases a residual error unit, and the situation can be effectively avoided by the shorcut structure inside.
In the technical scheme, ResNet is constructed by a plurality of residual error units, the residual error units can be regarded as the extension of two convolutional layers, short connection is added in the convolutional layers, and more sufficient characteristic information can be extracted. Further, the input value of the residual unit is X, and the output after the first layer linearization activation is f (X). After the second layer linear change, the short connection Identity is added, the total output of F (X) + X is obtained at this time, and the output value of the residual block is obtained finally after the activation function ReLU activation. In the residual unit, the improvement to the mapping function f (x) is simpler than the original mapping function h (x). H (X) can be considered as the sum of f (X) and the identity map X, and X for short connection paths neither increases the number of parameters nor affects the complexity of the original network model.
As a further improvement of the present invention, the construction of the early warning rule in step S105 specifically includes:
the first step is as follows: inputting historical data serving as a sample into a high-speed bearing temperature prediction model of a gearbox of the wind turbine generator system to obtain an acceleration bearing temperature predicted value, and subtracting the predicted value from a true value to obtain a temperature residual error;
the second step is that: and calculating the upper and lower limits of the residual error by using a 3-sigma criterion, evaluating the degree of the gear deviating from the normal state, and sending early warning information if the wind turbine generator gearbox exceeding the upper and lower limits of the residual error occurs n times or more in the second time frequency, wherein n is a natural number more than or equal to 3.
According to the technical scheme, multiple times of abnormity construction early warning and station real measurement analysis are utilized, compared with one time of abnormity construction early warning, the early warning accuracy is higher, meanwhile, whether early warning is conducted or not is judged according to a temperature residual error, compared with the early warning conducted by directly utilizing the temperature, the temperature residual error is that the degree of deviation of a gearbox from a normal state is evaluated through deviation between a predicted value and an actual value of the gearbox high-speed bearing temperature, and the judgment on the normal range of the gearbox high-speed bearing is more accurate.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
FIG. 1 is a flow chart of a wind turbine generator gearbox fault diagnosis analysis method based on a machine learning algorithm provided by the invention;
FIG. 2 is a flow chart in example 2 provided by the present invention;
FIG. 3 is a block diagram of a ResNet18 network according to embodiment 2 of the present invention;
FIG. 4 is a structural diagram of a residual error unit in embodiment 2 provided by the present invention;
fig. 5 is a schematic diagram of the KNN clustering algorithm provided by the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example 1
In this embodiment, a wind turbine generator gearbox fault diagnosis and analysis method based on a machine learning algorithm includes the following steps:
s101 sparse data construction: according to a set first time frequency, acquiring a plurality of types of operation parameters related to a gearbox of the wind turbine generator in the first time frequency from historical data and real-time data;
further, the S101 sparse data construction specifically includes: and taking 10min as a first time frequency, acquiring all characteristics of a wind field site as a plurality of types of operation parameters, wherein all characteristics comprise wind speed, wind power, gear box oil temperature, wind direction, yaw angle, pitch angle, voltage, current, instantaneous rotating speed of a generator, power grid active power, gear box intermediate shaft driving end bearing temperature, gear box intermediate shaft non-driving end bearing temperature, motor side gear box high speed shaft bearing temperature, gear box lubricating oil pool temperature, gear box lubricating oil inlet temperature and gear box oil way filter screen front oil pressure.
In this embodiment, 8-15min is used as a frequency, if the frequency is too low, detection is likely to be too frequent, data continuity is too strong, and the use is affected by overheating of the acquisition unit, and if the time is too long and is greater than 15min, data may not be acquired in time, and early warning is performed only when a problem occurs. Furthermore, the selected operation parameters directly influence the temperature of the rotating shaft of the motor.
The processing of the operation parameters specifically comprises the following steps: and taking the average value, the maximum value and the minimum value of each type of operation parameters in the first time frequency.
In the embodiment, if the correlation between all data characteristics and the high-speed bearing of the gearbox of the wind turbine generator is required to be obtained and the real-time data is used for analysis, the data volume is overlarge, and the efficiency is low. Sparse data with a frequency of 8-15 minutes is used here.
S102, data preprocessing: respectively carrying out null value processing and abnormal value processing on each type of operation parameters;
specifically, the null processing specifically includes: and sorting each type of operation parameters according to rows, directly adopting row deletion processing for continuously more than 5 null values in each row, and otherwise, replacing the null values with 0 values.
In this embodiment, since the evacuation data is controlled more, it is necessary to perform a separate null process, to ensure the purity of the data, and to reduce the impurity data.
Further, in the data preprocessing of step S102, the abnormal value processing specifically includes: and adopting a KNN clustering algorithm, taking K as a radius, wherein the K is a natural number more than or equal to 3, selecting K operation parameters closest to the operation parameters to be predicted for class distinguishing, and if the operation parameters to be predicted do not belong to any class of operation parameters, deleting the operation parameters as abnormal values.
In this embodiment, the general approximate value of K is 10, and outlier noise abnormal data with lower density in the original data is removed by using a KNN clustering algorithm. The basic principle of the algorithm is to assume that similar objects are very close to each other, the distance between the two objects is used for explaining the similarity, and the smaller the distance is, the higher the similarity is.
The abnormal value processing specifically comprises the following steps:
first step K value selection: taking historical data as sample data, splitting the sample data into training data and verification data according to a proportion, and selecting a proper K value through cross verification, wherein K is a natural number which is more than or equal to 3 and less than or equal to 30;
and a second step of data input: inputting a plurality of types of operation parameters in the historical data as tag data;
and thirdly, processing operation parameters: for each operation parameter x in each class, performing class judgment, specifically including:
3.1 distance calculation: calculating a distance data set q between the operation parameter x and other label data, wherein q is [ q ═ q [ [ q ] of any operation parameter x1,q2,...,qm]Wherein q1 represents the distance of the first label parameter from the operation parameter x, and qm represents the distance of the last label parameter from the operation parameter x;
3.2 sequencing: sequencing the obtained distance data sets q from small to large, and selecting the label data positioned at the first K;
3.3 classification: classifying the operation parameters x into A, wherein A is K label data and the operation parameter class with the most occurrence times;
and fourthly, processing abnormal data: and circularly performing the third step, taking the operation parameters which do not belong to any kind of operation parameters as abnormal data, and deleting the abnormal data.
The implementation of the KNN algorithm may be better understood with reference to the steps of outlier processing shown in FIG. 4. As shown in fig. 4, the square is the sample to be measured, and the five-pointed star and the triangle are the category 1 and the category 2. Judging samples belonging to the category 2, if the K value is 3, the range of the K value is a dotted line ellipse range, wherein the number of the category 2 is greater than that of the category 1, and the final result is judged to be the category 2; if the K value takes 5, the range taken by the K value is a solid line ellipse range, wherein the number of the categories 1 is greater than that of the categories 2, and the classification is finally determined as the category 2. Outliers far from the ellipse (with the K value as the radius) are not classified and are directly deleted.
In the embodiment, the abnormal value processing is realized through the classification, and the abnormal value processing performed by the clustering algorithm can effectively exclude the abnormal value. In this embodiment, what kind of (4) includes input characteristics such as wind speed, wind power, gearbox oil temperature, wind direction, yaw angle, pitch angle, voltage, and current.
S103, correlation analysis: selecting an operation parameter with higher correlation with a high-speed bearing of the gearbox from a plurality of types of operation parameters through a Pearson correlation coefficient;
the step S103 of correlation analysis includes the steps of: assuming two variables are present in the sample, then
Wherein x isiAnd yiThe values of variables x and y at time i,
and
is the average of x and y, respectively, xiGear box high speed bearing temperature at time i, yiSome kind of operating parameter, r, at time iiIs i time xiAnd y ofiAnd calculating the operation parameter with the correlation degree r larger than 0.6.
In the embodiment, the characteristics with high correlation with the high-speed bearing of the gearbox are selected through calculation of the correlation degree, and the characteristics are used as input of a wind turbine generator gearbox high-speed bearing temperature prediction model, so that the wind turbine generator gearbox high-speed bearing temperature can be more accurately fitted.
S104, constructing a high-speed bearing temperature prediction model of the wind turbine gearbox specifically comprises the following steps:
the first step is as follows: the operating parameters are used as input characteristic data, and the operation parameters are subjected to standardization processing, namely 0 mean value and 1 variance.
The second step is that: and constructing a Resnet18 network by using at least five convolutional layers, one maximum pooling layer and an average pooling layer, wherein in the Resnet18 network, the operation parameters after standardized processing are used as input, and an output value in a sigmod function form is obtained and used as a predicted value.
In this embodiment, a plurality of convolutional layers are provided to form a deeper network, so that finer functions can be captured and a better fitting effect can be obtained.
Preferably, the step S104 further includes adding a short connection, specifically: in the convolutional layers before the activation of the activation function ReLU, a short connection is inserted with the first and last convolutional layers as the connection.
In this embodiment, along with the increase of the number of network layers, there are also many problems, for example, information loss caused by the problem of gradient disappearance, and the gradient obtained by the activation output of each processing layer of the network is often small, so that propagation of the activation output and the gradient is poor, and the cost function calculation process is prolonged. This situation can lead to a rapid degradation of the accuracy of the convolutional neural network after optimization. Aiming at the problems, the temperature of the high-speed bearing of the gearbox of the wind generating set in the time period is predicted through the ResNet18 network, and the phenomenon of gradient disappearance caused by too deep network layers can be effectively avoided. In general neural networks, the deeper the network layer number is, the better the training effect is, but an over-fitting phenomenon and a gradient disappearance phenomenon occur when the depth is too deep, the over-fitting phenomenon is good in the training result, the testing result is poor, however, the resnet18 network increases a residual error unit, and the situation can be effectively avoided by the shorcut structure inside.
S104, construction of a wind turbine generator gearbox high-speed bearing temperature prediction model: constructing a high-speed bearing temperature prediction model of the wind turbine generator gearbox by utilizing a Resnet18 network and the operation parameters with higher correlation, and outputting a predicted value;
s105, constructing an early warning rule: and constructing an early warning rule for the deviation degree of the gearbox from a normal state by utilizing a residual between a predicted value and an actual value of the high-speed bearing temperature of the gearbox, and setting an upper residual limit value and a lower residual limit value according to a 3-sigma criterion to early warn the high-speed shaft temperature abnormality fault of the gearbox.
The early warning rule construction in the step S105 specifically comprises the following steps:
the first step is as follows: inputting historical data serving as a sample into a high-speed bearing temperature prediction model of a gearbox of the wind turbine generator system to obtain an acceleration bearing temperature predicted value, and subtracting the predicted value from a true value to obtain a temperature residual error;
the second step is that: and calculating the upper and lower limits of the residual error by using a 3-sigma criterion, evaluating the degree of the gear deviating from the normal state, and sending early warning information if the wind turbine generator gearbox exceeding the upper and lower limits of the residual error occurs n times or more in the second time frequency, wherein n is a natural number more than or equal to 3.
In the embodiment, multiple times of abnormity construction early warning and station real measurement analysis are utilized, compared with a single time of abnormity construction early warning, the early warning accuracy is higher, meanwhile, whether early warning is achieved or not is judged according to a temperature residual error, compared with the early warning which is directly carried out by utilizing the temperature, the temperature residual error means that the degree of deviation of a gearbox from a normal state is evaluated through deviation between a predicted value and an actual value of the temperature of a high-speed bearing of the gearbox, and the judgment on the normal range of the high-speed bearing of the gearbox is more accurate.
S106, early warning diagnosis analysis: and the staff performs early warning diagnosis analysis according to the fault early warning data and the field condition.
In the embodiment, sparse data is constructed firstly, and if real-time data is used for analysis to acquire correlation between all data characteristics and the high-speed bearing of the gearbox of the wind turbine generator, the data volume is too large, and the efficiency is low. Sparse data is selected here.
In the embodiment, correlation analysis is added, so that the operation parameters related to the core of the wind turbine gearbox can be screened out, redundant useless or low-correlation data are eliminated, operation is reduced, and early warning efficiency is improved.
In the embodiment, after the temperature of the high-speed bearing of the gearbox of the wind turbine generator is predicted, the residual error is compared with the actual value, and multiple comparisons are performed simultaneously, so that the early warning accuracy is higher, and the early warning caused by overhigh temperature in other special conditions is avoided.
In this embodiment, mainly be through the early warning of wind turbine generator system gear box high speed bearing temperature etc. its running state of direct reaction, for example the rotational speed is too high, perhaps transships, all can direct feedback to the temperature, and then through a data, the early warning of other relevant parameters can directly be carried out, the early warning of realization few data.
Example 2
Referring to fig. 2-4, shown are flow charts of a wind turbine generator gearbox fault diagnosis analysis method based on a machine learning algorithm according to an embodiment of the present invention, including:
s101: sparse data is constructed, actual operation data of a certain wind power plant collected by an SCADA system is adopted for analysis, all existing characteristics of the wind power plant are obtained by taking 8-15min as a first time frequency, and data types comprise wind speed, wind power, gear box oil temperature, wind direction, yaw angle, pitch angle, voltage, current and the like.
Ten-minute sparse data construction: data are collected every ten minutes, and the average value, the maximum value and the minimum value in the ten-minute variety of the relevant characteristics are taken. For example, the data is taken every ten minutes according to the characteristic of the temperature of the oil tank of the gearbox, and the average value, the maximum value and the minimum value of the data in ten minutes are obtained.
S102: data preprocessing: null and outlier processing methods.
And (4) null value processing: if the empty value in one row exceeds 5, the row change is directly deleted, otherwise, the empty value is replaced by a value of 0.
Processing abnormal values in historical and real-time data: and rejecting outlier noise abnormal data with lower density in the original data by adopting a KNN clustering algorithm. The basic principle of the algorithm is that similar objects are assumed to be very close to each other, the similarity of the two objects is described by using the distance between the two objects, the smaller the distance is, the higher the similarity is, the calculated similarity values are sorted, and K samples with the largest similarity values are selected for category judgment.
The first step is as follows: inputting all tag data y ═ y1,y2,...,ym]The fan characteristic x comprises wind speed, wind power, gear box oil temperature, wind direction, yaw angle, pitch angle, voltage, current and the like, and the number K of the approaching points is set.
The second step is that: for each fan characteristic x:
(1) calculating the distance q between the low-dimensional discrimination feature and all the label data as q1,q2,...,qm]；
(2) Calculating q ═ q1,q2,...,qm]Arranging according to the sequence from small to large;
(3) taking out the first 10 with the minimum distance, and finding out the label data corresponding to the distance to classify;
(4) and judging the type of the low-dimensional distinguishing features according to a certain decision scheme.
Discrete anomaly data can be identified using a KNN clustering algorithm by adjusting appropriate parameters.
S103: and (4) performing correlation analysis, and selecting a characteristic with high correlation with the high-speed bearing of the gearbox in the data through a Pearson correlation coefficient.
And selecting the characteristic with high correlation with the high-speed bearing of the gearbox in the data after data preprocessing through the Pearson correlation coefficient, and selecting the characteristic with the correlation coefficient larger than 0.6 as the input of the fault prediction model.
The correlation coefficient is a measure for measuring the degree of correlation between two variables in the data set. The pearson correlation coefficient is one of phase coefficients, and is mainly used to describe the magnitude of linear correlation of two random variables in a sample set. Assuming that there are two variables in the sample, then
Wherein x isiAnd yiThe values of variables x and y at time i,
and
the average values of x and y, respectively. E.g. calculating the degree of correlation, x, between the gearbox high speed bearing temperature and a certain fan characteristiciGear box high speed bearing temperature at time i, yiThe degree of correlation between certain fan characteristics (e.g. wind speed) at time i,
in order to maintain the high-speed bearing temperature of the gearbox,
is a certain fan characteristic.
The characteristics which are selected by calculation and have the temperature dependency of the high-speed bearing of the gear box larger than 0.6 comprise the following components: instantaneous rotating speed of a generator, active power of a power grid, bearing temperature of a drive end bearing of a middle shaft of a gear box, bearing temperature of a non-drive end bearing of the middle shaft of the gear box, bearing temperature of a high-speed shaft of a gear box on the side of a motor, lubricating oil pool temperature of the gear box, lubricating oil inlet temperature of the gear box and oil pressure in front of an oil way filter screen of the gear box.
Specifically, the feature with the larger correlation is included in the full-amount data features selected in s101, and the feature with the correlation larger than 0.6 in all data of s101 is selected according to the correlation analysis.
S104: and (3) constructing a high-speed bearing temperature prediction model of the wind turbine generator gearbox.
The first step is as follows: input feature data were normalized, 0 mean, 1 variance.
The second step is that: generally, a deeper network can capture finer functions and obtain better fitting effect. However, along with the increase of the number of network layers, there are many problems such as information loss caused by the gradient disappearance problem, and the gradient obtained by the activation output of each processing layer of the network is often small, so that the propagation of the activation output and the gradient is poor, and the cost function calculation process is prolonged. This situation can lead to a rapid degradation of the accuracy of the convolutional neural network after optimization. Aiming at the problems, the temperature of the high-speed bearing of the gearbox of the wind generating set in the time period is predicted through the ResNet18 network, and the phenomenon of gradient disappearance caused by too deep network layers can be effectively avoided.
ResNet is constructed by a plurality of residual error units, the residual error units can be regarded as the extension of two convolutional layers, short connection is added in the convolutional layers, and more sufficient characteristic information can be extracted. As shown in fig. 4, the input value of the residual unit is X, and the output after the first-layer linearization activation is f (X). After the second layer linear change, the short connection Identity is added, the total output of F (X) + X is obtained at this time, and the output value of the residual block is obtained finally after the activation function ReLU activation. In the residual unit, the improvement to the mapping function f (x) is simpler than the original mapping function h (x).
H (X) can be considered as the sum of f (X) and the identity map X, and X for short connection paths neither increases the number of parameters nor affects the complexity of the original network model.
The specific parameters of the ResNet18 network are shown in table 1:
TABLE 1 ResNet18 network parameters
S105: and (3) constructing an early warning rule, namely constructing the early warning rule for the deviation degree of the gearbox from the normal state by utilizing the residual between the predicted value and the actual value of the high-speed bearing temperature of the gearbox, and setting the upper and lower limit values of the residual according to a 3-sigma rule to early warn the abnormal temperature fault of the high-speed shaft of the gearbox. 3-sigma represents that 99.7% of data is in the specification, so the defect is only 0.3%, and in the past, 3-sigma is regarded as a qualified standard level, and the 3-sigma is more accurate in defining the normal range of the residual error.
The first step is as follows: and subtracting the obtained predicted value and the true value of the acceleration bearing temperature to obtain a temperature residual error.
The second step is that: and calculating the upper and lower limits of the residual error by using a 3-sigma rule, evaluating the degree of the gear deviating from the normal state, and sending out early warning information if the upper and lower limits of the residual error exceed 3 times in a set within one hour.
S106: and (4) checking and analyzing the early warning condition, namely, measuring the temperature of the actual part by wind field workers, and analyzing the temperature in combination with early warning to judge the authenticity of the fault early warning.
According to the invention, the high-speed bearing of the gear box is in a high-speed, high-temperature and variable-load working environment for a long time, and the operation and maintenance difficulty is higher. Therefore, the fault diagnosis of the high-speed bearing of the gearbox of the wind turbine generator is very necessary to ensure the safe and stable operation of the equipment. The method has strong practicability, realizes accurate prediction of temperature abnormity of the high-speed bearing of the gear box, effectively avoids the condition that the production efficiency of equipment is influenced and even the whole unit is stopped and maintained due to the fault of the high-speed bearing of the gear box, and has great significance.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A wind turbine generator gearbox fault diagnosis and analysis method based on a machine learning algorithm is characterized by comprising the following steps:
s101 sparse data construction: according to a set first time frequency, acquiring a plurality of types of operation parameters related to a gearbox of the wind turbine generator in the first time frequency from historical data and real-time data;
s102, data preprocessing: respectively carrying out null value processing and abnormal value processing on each type of operation parameters;
s103, correlation analysis: selecting an operation parameter with high correlation with a high-speed bearing of the gearbox from a plurality of types of operation parameters through a Pearson correlation coefficient;
s104, construction of a wind turbine generator gearbox high-speed bearing temperature prediction model: constructing a high-speed bearing temperature prediction model of the wind turbine generator gearbox by utilizing a Resnet18 network and the operation parameters with higher correlation, and outputting a predicted value;
s105, constructing an early warning rule: constructing an early warning rule for the deviation degree of the gearbox from a normal state by utilizing a residual between a predicted value and an actual value of the high-speed bearing temperature of the gearbox, and setting upper and lower limit values of the residual according to a 3-sigma criterion to early warn the fault of the high-speed shaft temperature abnormality of the gearbox;
s106, early warning diagnosis analysis: and the staff performs early warning diagnosis analysis according to the fault early warning data and the field condition.
2. The wind turbine generator gearbox fault diagnosis and analysis method based on the machine learning algorithm as claimed in claim 1, wherein the S101 sparse data construction specifically comprises: and taking 8-15min as a first time frequency, acquiring all characteristics of a wind field site as a plurality of types of operation parameters, wherein all characteristics comprise wind speed, wind power, gearbox oil temperature, wind direction, yaw angle, pitch angle, voltage, current, instantaneous rotating speed of a generator, power grid active power, bearing temperature of a drive end bearing of a gearbox intermediate shaft, bearing temperature of a non-drive end bearing of the gearbox intermediate shaft, bearing temperature of a high-speed shaft of a motor-side gearbox, lubricating oil pool temperature of the gearbox, lubricating oil inlet temperature of the gearbox and oil pressure in front of an oil way filter screen of the gearbox.
3. The wind turbine generator gearbox fault diagnosis and analysis method based on the machine learning algorithm as claimed in claim 1, wherein after the sparse data is constructed in step S101, the operation parameter processing is further included before the data preprocessing in step S102, specifically: and taking the average value, the maximum value and the minimum value of each type of operation parameters in the first time frequency.
4. The wind turbine generator gearbox fault diagnosis and analysis method based on the machine learning algorithm as claimed in claim 1, wherein in the step S102 data preprocessing, the null value processing specifically comprises: and sorting each type of operation parameters according to rows, directly adopting row deletion processing for continuously more than 5 null values in each row, and otherwise, replacing the null values with 0 values.
5. The wind turbine generator gearbox fault diagnosis and analysis method based on the machine learning algorithm as claimed in claim 1, wherein in the data preprocessing of step S102, a KNN clustering algorithm is adopted in the abnormal value processing, K is used as a radius, K is a natural number greater than or equal to 3, K operation parameters closest to the operation parameter to be predicted are selected for class discrimination, and if the operation parameter to be predicted does not belong to any one of the operation parameters, the operation parameter to be predicted is used as an abnormal value to be deleted.
6. The wind turbine generator gearbox fault diagnosis analysis method based on the machine learning algorithm as claimed in claim 5, wherein the abnormal value processing specifically comprises the following steps:
first step K value selection: taking historical data as sample data, splitting the sample data into training data and verification data according to a proportion, and selecting a proper K value through cross verification, wherein K is a natural number which is more than or equal to 3 and less than or equal to 30;
and a second step of data input: inputting a plurality of types of operation parameters in the historical data as tag data;
and thirdly, processing operation parameters: for each operation parameter x in each class, performing class judgment, specifically including:
3.1 distance calculation: calculating a distance data set q between the operation parameter x and other label data, wherein q is [ q ═ q [ [ q ] of any operation parameter x1,q2,...,qm]Wherein q1 represents the distance between the first label parameter and the operation parameter x, and qm represents the distance between the last label parameter and the operation parameter x;
3.2 sequencing: sequencing the obtained distance data sets q from small to large, and selecting the label data positioned at the first K;
3.3 classification: classifying the operation parameters x into A, wherein A is K label data and the operation parameter class with the most occurrence times;
and fourthly, processing abnormal data: and circularly performing the third step, taking the operation parameters which do not belong to any kind of operation parameters as abnormal data, and deleting the abnormal data.
7. The wind turbine generator gearbox fault diagnosis analysis method based on the machine learning algorithm as claimed in claim 1, wherein the step S103 of correlation analysis comprises the following steps: assuming two variables are present in the sample, then
Wherein x isiAnd yiThe values of variables x and y at time i,
and
is the average of x and y, respectively, xiGear box high speed bearing temperature at time i, yiA certain class of operating parameters at time i, riIs i time xiAnd y ofiAnd calculating the operation parameter with the correlation degree r larger than 0.6.
8. The wind turbine generator gearbox fault diagnosis analysis method based on the machine learning algorithm as claimed in claim 1, wherein the S104 wind turbine generator gearbox high speed bearing temperature prediction model is specifically constructed as follows:
the first step is as follows: the operating parameters are used as input characteristic data, and the operation parameters are subjected to standardization processing, namely 0 mean value and 1 variance.
The second step is that: and constructing a Resnet18 network by using at least five convolutional layers, one maximum pooling layer and an average pooling layer, wherein in the Resnet18 network, the operation parameters after standardized processing are used as input, and an output value in a sigmod function form is obtained and used as a predicted value.
9. The wind turbine generator gearbox fault diagnosis and analysis method based on the machine learning algorithm as claimed in claim 8, wherein the step S104 further includes an increase of short connections, specifically: in the convolutional layers before the activation of the activation function ReLU, a short connection is inserted with the first and last convolutional layers as the connection.
10. The wind turbine generator gearbox fault diagnosis and analysis method based on the machine learning algorithm as claimed in claim 1, wherein the early warning rule construction in step S105 specifically comprises:
the first step is as follows: inputting historical data serving as a sample into a high-speed bearing temperature prediction model of a gearbox of the wind turbine generator system to obtain an acceleration bearing temperature predicted value, and subtracting the predicted value from a true value to obtain a temperature residual error;
the second step is that: and calculating the upper and lower limits of the residual error by using a 3-sigma criterion, evaluating the degree of the gear deviating from the normal state, and sending early warning information if the wind turbine generator gearbox exceeding the upper and lower limits of the residual error occurs n times or more in the second time frequency, wherein n is a natural number more than or equal to 3.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111501998 | 2021-12-09 | ||
| CN202111501998X | 2021-12-09 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114444382A true CN114444382A (en) | 2022-05-06 |
| CN114444382B CN114444382B (en) | 2024-07-19 |
Family
ID=
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115358281A (en) * | 2022-10-21 | 2022-11-18 | 深圳市耐思特实业有限公司 | Machine learning-based cold and hot all-in-one machine monitoring control method and system |
| CN115479636A (en) * | 2022-10-26 | 2022-12-16 | 中国航空工业集团公司金城南京机电液压工程研究中心 | Turbine cooler state monitoring and analyzing method |
| CN115795351A (en) * | 2023-01-29 | 2023-03-14 | 杭州市特种设备检测研究院(杭州市特种设备应急处置中心) | Elevator big data risk early warning method based on residual error network and 2D feature representation |
| CN116401536A (en) * | 2023-06-08 | 2023-07-07 | 华电电力科学研究院有限公司 | Fault diagnosis method, device and system for anemometer of wind turbine generator |
| CN117171693A (en) * | 2023-10-30 | 2023-12-05 | 山东交通学院 | Cutting abnormality detection method in woodworking polishing process |
| CN117605627A (en) * | 2024-01-22 | 2024-02-27 | 南京讯联液压技术股份有限公司 | System and method for detecting oil leakage of core body of wind power gear box cooler |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110907066A (en) * | 2019-11-30 | 2020-03-24 | 华能如东八仙角海上风力发电有限责任公司 | Wind turbine generator gearbox bearing temperature state monitoring method based on deep learning model |
| CN113240016A (en) * | 2021-05-18 | 2021-08-10 | 哈尔滨理工大学 | Rolling bearing fault diagnosis method based on deep learning |
| CN113343855A (en) * | 2021-06-09 | 2021-09-03 | 西南交通大学 | Rolling bearing fault diagnosis system and method based on guide type sub-field self-adaption |
| CN113361186A (en) * | 2021-04-28 | 2021-09-07 | 山东大学 | Complete data-based wind turbine generator fault diagnosis method |
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110907066A (en) * | 2019-11-30 | 2020-03-24 | 华能如东八仙角海上风力发电有限责任公司 | Wind turbine generator gearbox bearing temperature state monitoring method based on deep learning model |
| CN113361186A (en) * | 2021-04-28 | 2021-09-07 | 山东大学 | Complete data-based wind turbine generator fault diagnosis method |
| CN113240016A (en) * | 2021-05-18 | 2021-08-10 | 哈尔滨理工大学 | Rolling bearing fault diagnosis method based on deep learning |
| CN113343855A (en) * | 2021-06-09 | 2021-09-03 | 西南交通大学 | Rolling bearing fault diagnosis system and method based on guide type sub-field self-adaption |
Non-Patent Citations (1)
| Title |
|---|
| 王灿;李韶武;张天阳;吕微;: "风电机组齿轮箱油温异常诊断方法与预警策略研究", 船舶工程, no. 1, 15 July 2020 (2020-07-15), pages 542 - 546 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115358281A (en) * | 2022-10-21 | 2022-11-18 | 深圳市耐思特实业有限公司 | Machine learning-based cold and hot all-in-one machine monitoring control method and system |
| CN115479636A (en) * | 2022-10-26 | 2022-12-16 | 中国航空工业集团公司金城南京机电液压工程研究中心 | Turbine cooler state monitoring and analyzing method |
| CN115479636B (en) * | 2022-10-26 | 2023-10-03 | 中国航空工业集团公司金城南京机电液压工程研究中心 | Turbine cooler state monitoring and analyzing method |
| CN115795351A (en) * | 2023-01-29 | 2023-03-14 | 杭州市特种设备检测研究院(杭州市特种设备应急处置中心) | Elevator big data risk early warning method based on residual error network and 2D feature representation |
| CN116401536A (en) * | 2023-06-08 | 2023-07-07 | 华电电力科学研究院有限公司 | Fault diagnosis method, device and system for anemometer of wind turbine generator |
| CN116401536B (en) * | 2023-06-08 | 2023-08-15 | 华电电力科学研究院有限公司 | Fault diagnosis method, device and system for anemometer of wind turbine generator |
| CN117171693A (en) * | 2023-10-30 | 2023-12-05 | 山东交通学院 | Cutting abnormality detection method in woodworking polishing process |
| CN117171693B (en) * | 2023-10-30 | 2024-01-26 | 山东交通学院 | Cutting abnormality detection method in woodworking polishing process |
| CN117605627A (en) * | 2024-01-22 | 2024-02-27 | 南京讯联液压技术股份有限公司 | System and method for detecting oil leakage of core body of wind power gear box cooler |
| CN117605627B (en) * | 2024-01-22 | 2024-03-19 | 南京讯联液压技术股份有限公司 | System and method for detecting oil leakage of core body of wind power gear box cooler |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111539553B (en) | Wind turbine generator fault early warning method based on SVR algorithm and off-peak degree | |
| CN110410282B (en) | SOM-MQE and SFCM (Small form-factor pluggable) based wind turbine generator health state online monitoring and fault diagnosis method | |
| CN108412710B (en) | A kind of Wind turbines wind power data cleaning method | |
| CN111537219B (en) | Fan gearbox performance detection and health assessment method based on temperature parameters | |
| CN109543210A (en) | A kind of Wind turbines failure prediction system based on machine learning algorithm platform | |
| CN105677791B (en) | For analyzing the method and system of the operation data of wind power generating set | |
| CN110926809B (en) | Big data analysis-based wind turbine generator transmission chain fault early warning method | |
| CN110362045B (en) | Marine doubly-fed wind turbine generator fault discrimination method considering marine meteorological factors | |
| CN111878320B (en) | Monitoring method and system of wind generating set and computer readable storage medium | |
| CN107728059B (en) | Pitch system state evaluation method | |
| CN109887122A (en) | A kind of ship lock functional failure of electromechanical long-distance intelligent diagnostic method and system | |
| CN113339204B (en) | Wind driven generator fault identification method based on hybrid neural network | |
| CN112065668A (en) | Wind turbine generator state abnormity assessment method and system | |
| CN115614292B (en) | Vibration monitoring device and method for vertical water pump unit | |
| Shi et al. | Study of wind turbine fault diagnosis and early warning based on SCADA data | |
| CN116956215A (en) | Fault diagnosis method and system for transmission system | |
| CN114169718A (en) | Method for improving reliability of wind turbine generator based on state evaluation of wind turbine generator | |
| CN114463143A (en) | Method for enhancing SCADA fault data of offshore doubly-fed wind generator | |
| Elijorde et al. | A wind turbine fault detection approach based on cluster analysis and frequent pattern mining | |
| CN116910128A (en) | Wind turbine generator hard and soft fault data analysis system and analysis method thereof | |
| CN114444382B (en) | Wind turbine generator system gear box fault diagnosis and analysis method based on machine learning algorithm | |
| CN114444382A (en) | Wind turbine generator gearbox fault diagnosis and analysis method based on machine learning algorithm | |
| CN115310746A (en) | State evaluation method and system for main transmission system of wind generating set | |
| Zhou et al. | Abnormal data processing of wind turbine based on combined algorithm and class center imputation | |
| CN114154567A (en) | Wind power plant station operation data anomaly identification method based on machine learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |