CN111814849B - DA-RNN-based wind turbine generator set key component fault early warning method - Google Patents
DA-RNN-based wind turbine generator set key component fault early warning method Download PDFInfo
- Publication number
- CN111814849B CN111814849B CN202010573207.3A CN202010573207A CN111814849B CN 111814849 B CN111814849 B CN 111814849B CN 202010573207 A CN202010573207 A CN 202010573207A CN 111814849 B CN111814849 B CN 111814849B
- Authority
- CN
- China
- Prior art keywords
- value
- early warning
- variable
- data
- wind turbine
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 42
- 230000007246 mechanism Effects 0.000 claims abstract description 27
- 238000007781 pre-processing Methods 0.000 claims abstract description 20
- 238000012423 maintenance Methods 0.000 claims abstract description 19
- 238000013528 artificial neural network Methods 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims description 42
- 239000013598 vector Substances 0.000 claims description 31
- 230000002159 abnormal effect Effects 0.000 claims description 18
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000012886 linear function Methods 0.000 claims description 6
- 238000012417 linear regression Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 230000005856 abnormality Effects 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 3
- 230000001568 sexual effect Effects 0.000 claims 1
- 238000013461 design Methods 0.000 abstract description 5
- 238000012544 monitoring process Methods 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 9
- 238000012360 testing method Methods 0.000 description 8
- 230000008901 benefit Effects 0.000 description 4
- 238000012790 confirmation Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000013101 initial test Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Wind Motors (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses a method for early warning faults of key components of a wind turbine generator based on a circulating neural network DA-RNN with a double-attention mechanism. The method is based on a data acquisition and monitoring control (SCADA) system data set collected in a normal running state of a wind turbine, designs a preprocessing flow, selects a DA-RNN model for real-time variable estimation, designs through multi-threshold setting and judgment criteria, outputs a judgment result sequence, and gives out a final early warning result according to the judgment result sequence. In the fault early warning method, a preprocessing flow is designed aiming at different types of noise data, and a reliable data base is provided; the DA-RNN model comprehensively considers the influence of related variables and historical information, and distributes different weights, so that the accuracy of variable estimation is ensured; the multi-threshold setting and the judgment criterion design avoid single 0-1 judgment, so that the final early warning result has robustness; finally, the fault early warning of the key components is realized, the downtime of the unit is reduced, the operation and maintenance cost is saved, and the method has stronger theories and practicability.
Description
Technical Field
The invention relates to a fault early warning method for a key component of a wind turbine generator based on DA-RNN (data-random number network), which is a method for carrying out fault early warning on the key component by designing a normalized data preprocessing flow based on a normal running state data set of a fan, selecting DA-RNN as a variable estimation model to carry out real-time estimation on a target variable and designing a fault early warning strategy selected by combining multi-threshold setting with multi-discriminant criteria based on real-time running residual errors.
Background
With the increasing scarcity of global pollution and traditional fossil energy, the development of clean energy attracts a great deal of attention, wind energy is rapidly developed with the advantage of clean and pollution-free wind energy, and the wind power industry is one of novel renewable energy industries which are greatly developed at home and abroad. At present, the total capacity of the fan installed in China is in the front of the world, but the rapid development of the wind power generation market in recent years also leads to insufficient preparation in the research and development period, and the operation and maintenance cost of the fan is high.
The high failure rate of the fan is a main factor causing high operation and maintenance cost, the wind turbine generator is a complex system consisting of a plurality of components and subsystems, the wind turbine generator usually operates in remote areas such as suburb plain, mountain area, coast, and the like, the operating environment is severe and changeable, and the failure of key components can cause the shutdown and maintenance of the whole machine, so that a large amount of economic losses are brought. Therefore, the early identification of the abnormality of the key component is realized, the early abnormality is prevented from evolving into a catastrophic failure, the failure early warning of the key component is realized, and the predictive maintenance is performed, so that the method has great significance in reducing the operation and maintenance cost and realizing the intelligent operation and maintenance of the wind power plant. However, the existing variable estimation model for fault early warning is difficult to comprehensively consider the influence of related variables and historical information, and the existing simple early warning strategy is also difficult to ensure the accuracy of early warning results. Therefore, a more accurate variable estimation model is selected, and a robust early warning strategy is designed, so that the method has great significance in realizing accurate fault early warning.
Disclosure of Invention
The invention aims to provide a method for early warning faults of key components of a wind turbine generator based on DA-RNN by accurately estimating target variables and designing a robust early warning strategy to early warn faults of the key components of the wind turbine generator. According to the method, a data set of a normal running state of the wind turbine generator is selected, firstly, different noise data types are considered to design a data preprocessing flow, then DA-RNN is selected as a variable estimation model, the influences of related variables and historical information are comprehensively considered to estimate a target variable in real time, the accuracy of estimation is guaranteed, a robust early warning strategy is designed, different early warning result requirements and different abnormal characteristics are considered, and the flexibility and the accuracy of an early warning result are guaranteed. The method can be extended to each key component with a temperature measuring point of the wind turbine generator, fault early warning of the key component is achieved, and the method has practical value and high expansibility.
The aim of the invention is achieved by the following technical scheme: a wind turbine generator system key component fault early warning method based on DA-RNN includes the following steps:
1) Selecting a wind turbine to be subjected to fault early warning, acquiring N pieces of operation data recorded in a SCADA system of the wind turbine under a normal operation state, wherein key components of the wind turbine comprise a gear box, a generator, a variable pitch system and the like, selecting a temperature variable measured at a temperature measuring point of a component to be early warned in the SCADA system as a target variable y, and taking all variables related to the temperature of the component as related variables X to construct an initial training set
2) Offline training stage, based on initial training setDesigning a data preprocessing flow, wherein the preprocessing step comprises the steps of eliminating and interpolating isolated abnormal points, eliminating and interpolating data during fan shutdown maintenance based on operation and maintenance records, interpolating missing values, and preprocessing a training set [ X ] train ,y train ]Performing model training as input of a variable estimation model;
3) Selecting a circulating neural network DA-RNN model based on a double-attention mechanism as a variable estimation model, and selectingThe length W of the sliding window is represented as t at the moment in the sliding window w W=t-w+1, T-w+2, &..t, where T is the current time, each time T is assigned by designing an input attention mechanism w Related variableFor the target variable->Influence weight of->Reconstruction related variable +.>As encoder input, the encoder part is a plurality of LSTM units, the input of each LSTM unit is a reconstruction related variable at one moment in the sliding window, and the encoder part outputs as a hidden vector h; then designing a time attention mechanism to distribute hidden vectors of different historical moments in the sliding window to target variables of the current moment +.>Influence weight of->Obtaining a semantic vector c, integrating the semantic vector and a target variable history value by using a linear regression model as input of a decoder, wherein the decoder is divided into a plurality of LSTM units and outputs the LSTM units as a time t before the current time T-1 Obtaining the estimated value of the target variable at the current moment through a linear function>
4) Subtracting the model output at the corresponding moment from the actual value of the target variable in the training set, namely the estimated value of the target variable, obtaining an estimated residual sequence of the training set, and obtaining the average value mu of the residual sequence train Standard deviation sigma train ；
5) Based on trainingThe set residual sequence is subjected to multi-threshold setting, wherein the multi-threshold setting is that the average value mu of the training set estimated residual sequence train Plus or minus k times standard deviation sigma train Respectively used as the upper and lower limits of the residual sequence threshold, wherein the upper limit U r (k)＝μ train +kσ train Lower limit L r (k)＝μ train -kσ train The method comprises the steps of carrying out a first treatment on the surface of the The higher the upper limit of the threshold value is, or the lower limit is, the fewer the number of data points exceeding the threshold value is in the online application stage, the false alarm rate in the corresponding fault early warning result is reduced, and the false alarm rate is increased;
6) In the online application stage, based on real-time operation data point d and on DA-RNN model trained in the offline stage, an estimated residual value r of the actual measured value of the data point d minus the estimated value of the model is obtained d ；
7) Selecting a value of k, determining upper and lower thresholds, if r d Exceeding the corresponding threshold of the current k value, i.e. greater than the upper threshold limit U r (k) Or less than the threshold lower limit L r (k) Calculating the number of data points count1 (k) which continuously exceed the threshold value before the data point d, and calculating the percentage value count2 (k)% of the number of data points which exceed the threshold value in the time range of day before the data point d, if r d At the upper threshold limit U r (k) And a lower threshold limit L r (k) Outputting two judgment results 0;
8) Setting multiple discriminants, wherein the multiple discriminants are continuous overrun discriminants combined with percentage overrun discriminants, setting a continuous overrun discriminant threshold parameter S (k) and a percentage overrun discriminant threshold parameter P (k)%, recording the judgment result of the k-S (k) early warning strategy combination as 1 when the condition count1 (k) is more than or equal to S (k) is met, recording 0 when the condition count2 (k) is not met, recording 1 when the condition count2 (k) is more than or equal to P (k) is met, and recording 0 when the condition is not met;
9) Repeating the step 7) and the step 8) for all other values of k to obtain a judging result sequence, and judging whether a final alarm is given for the real-time data point d based on the judging result sequence.
Further, in the step 2), the pretreatment flow of the offline training includes the following steps:
a) The isolated abnormal point is usually a recording error caused by sensor abnormality, and is judged by an operation mechanism, and the judgment conditions are as follows: for temperature variables, the value is greater than 150 degrees or less than 0 degrees; for the wind speed variable, the numerical value is larger than the cut-out wind speed of the unit or smaller than the cut-in wind speed; for the power variable, the numerical value is larger than the rated power of the unit or is a negative value; when the conditions are met, the data are judged to be isolated abnormal points and are removed;
b) When the wind turbine generator is in an operation, maintenance and overhaul period, the wind turbine generator is in a shutdown state, and the data recorded by the SCADA system is usually 0 value or a system default value, wherein the value cannot represent the normal operation state of the wind turbine generator, so that the data in the shutdown, maintenance and overhaul period are removed;
c) In order to ensure the time continuity of the data, the missing values in the removed data and the SCADA system are interpolated, wherein the interpolation method is mean value interpolation, namely, the mean value of the data 1 hour before the variable interpolation position is taken as the interpolation value at the current moment.
Further, in the step 3), the model is input as a sliding window related variableWherein n is the number of related variables, the input attention mechanism part inputs the related variables X, the hidden state output h and the memory unit output s of the encoder, and the influence weight of the kth related variable on the target variable is output->The calculation process is as follows:
wherein v is en ,W en ,U en For the parameter to be learned, at time t w Based on impact weightsReconstructing the related variable to obtain a time t w Is expressed as:
the reconstructed vector is input as an encoder part, the encoder is composed of a plurality of LSTM units, and the reconstructed vector is output as a hidden state h and a memory unit s, at a time t w The update of LSTM is performed by forget gate f t Input gate i t Output gate o t The decision, its update rule is as follows:
wherein W is f ,W i ,W o ,W s ,b f ,b i ,b o ,b s For the parameter to be learned, σ is a sigmoid function, and the corresponding elements are multiplied by the ";
the input of the time attention mechanism is a hidden state h, a decoder hidden state output h 'and a memory unit output s', and the weight of the ith hidden state is outputThe calculation process is as follows:
wherein v is de ,W de ,U de Reconstructing the hidden state based on the weight beta to obtain a semantic vector for the parameter to be learned, wherein the semantic vector is expressed as follows:
then, integrating the semantic vector and the target variable historical value by using a linear regression model to obtain the input of the decoderExpressed as:
wherein the method comprises the steps ofIs a parameter to be learned;
the decoder is also composed of LSTM units, forgetting the gate f t ' input gate i t ' output door o t The update rule of' is as follows:
wherein W' f ,W′ i ,W′ o ,W′ s ,b′ f ,b′ i ,b′ o ,b′ s Is a parameter to be learned;
finally, obtaining the estimated value of the T target variable at the current moment through a linear functionThe following is shown:
wherein W is T ,b T Is a parameter to be learned.
Further, in the step 5), it may be assumed that the training set residual sequence is subject to normal distribution, and the value of k in the multi-threshold setting is [1.5,2,2.5,3] based on the normal distribution characteristic.
Further, in the step 8), the multiple criteria correspond to different abnormal characteristics, the continuous overrun criterion corresponds to abnormal characteristics with continuous abnormal values, the percentage overrun criterion corresponds to abnormal conditions with large data fluctuation, and the continuous overrun criterionCriterion threshold parameter S k Threshold parameter P of percentage overrun criterion k % is set asWherein C is a constant.
Further, in the step 9), for the real-time data point d, two 0-1 judgment results are output for each k value, and finally, the output is a 0-1 sequence, the dimension is 2k dimension, when the number of 1 in the sequence is greater than k, an alarm is given to the real-time data point, and when the number of 1 in the sequence is less than or equal to k, no alarm is given.
Compared with the prior art, the invention has the following innovative advantages and remarkable effects:
1) The normalized data preprocessing flow is designed aiming at different types of noise data existing in the data set in the normal operation period, so that the constructed training set can accurately represent the normal operation state of the wind turbine generator, and a data base is provided for accurate estimation of target variables;
2) The DA-RNN is selected as a variable estimation model, influences of related variables and historical information are comprehensively considered, different influence weights on target variable estimation are determined through a double-layer attention mechanism, and reliability and accuracy of the model are guaranteed;
3) The method has the advantages that a robust fault early warning strategy is designed, the requirements of early warning results on different false alarm rates and false alarm rates can be comprehensively considered through multi-threshold setting, two discriminant criteria perform more comprehensive early warning on different abnormal characteristics, a final early warning result is given out based on a judging result sequence, and the combination of different early warning strategies is comprehensively considered, so that the robustness and the accuracy of the early warning result are ensured;
4) The invention relates to a critical component fault early warning method aiming at temperature parameters, which is applicable to all the critical components of a wind turbine generator set with corresponding temperature measuring points and has expansibility.
Drawings
FIG. 1 is a flow chart of a fault early warning method for key components of a wind turbine generator;
FIG. 2 is a diagram of data prior to preprocessing of selected target variables in an embodiment of the present invention;
FIG. 3 is a diagram of data after preprocessing of selected target variables in an embodiment of the invention;
FIG. 4 is a schematic diagram of an encoder in which a selected DA-RNN model of the present invention incorporates an input attention mechanism;
FIG. 5 is a schematic diagram of a decoder incorporating a time attention mechanism for the selected DA-RNN model of the present invention;
FIG. 6 is a graph of the target variable estimation results of the variable estimation model according to an embodiment of the present invention;
FIG. 7 is a schematic diagram of a multi-threshold setting of an embodiment of the present invention;
fig. 8 is a diagram of early warning results according to an embodiment of the present invention.
Detailed Description
In order that the above objects, features and advantages of the invention will be readily understood, a more particular description of the invention will be rendered by reference to the appended drawings.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways other than those described herein, and persons skilled in the art will readily appreciate that the present invention is not limited to the specific embodiments disclosed below.
As shown in fig. 1, the method for early warning faults of key components of a wind turbine generator set based on DA-RNN provided by the present application includes:
1) Selecting a wind turbine to be subjected to fault early warning, acquiring N pieces of operation data recorded in a SCADA system of the wind turbine under a normal operation state, wherein key components of the wind turbine comprise a gear box, a generator, a variable pitch system and the like, selecting a temperature variable measured at a temperature measuring point of a component to be early warned in the SCADA system as a target variable y, and taking all variables related to the temperature of the component as related variables X to construct an initial training set
2) Offline training stage, based on initial training setDesigning a data preprocessing flow, wherein the preprocessing step comprises the steps of eliminating and interpolating isolated abnormal points, eliminating and interpolating data during fan shutdown maintenance based on operation and maintenance records, interpolating missing values, and preprocessing a training set [ X ] train ,y train ]Performing model training as input of a variable estimation model;
3) Selecting a circulating neural network DA-RNN model based on a double-attention mechanism as a variable estimation model, selecting a sliding window length W, and expressing the moment in the sliding window as t w W=t-w+1, T-w+2, &..t, where T is the current time, each time T is assigned by designing an input attention mechanism w Related variableFor the target variable->Influence weight of->Reconstruction related variable +.>As encoder input, the encoder part is a plurality of LSTM units, the input of each LSTM unit is a reconstruction related variable at one moment in the sliding window, and the encoder part outputs as a hidden vector h; then designing a time attention mechanism to distribute hidden vectors of different historical moments in the sliding window to target variables of the current moment +.>Influence weight of->Obtaining a semantic vector c, integrating the semantic vector and a target variable history value by using a linear regression model as input of a decoder, wherein the decoder is divided into a plurality of LSTM units and outputs the LSTM units as a time t before the current time T-1 Obtaining the estimated value of the target variable at the current moment through a linear function>
4) Subtracting the model output at the corresponding moment from the actual value of the target variable in the training set, namely the estimated value of the target variable, obtaining an estimated residual sequence of the training set, and obtaining the average value mu of the residual sequence train Standard deviation sigma train ；
5) Multi-threshold setting is carried out based on the residual sequence of the training set, and the multi-threshold setting is that the average value mu of the estimated residual sequence of the training set is set train Plus or minus k times standard deviation sigma train Respectively used as the upper and lower limits of the residual sequence threshold, wherein the upper limit U r (k)＝μ train +kσ train Lower limit L r (k)＝μ train -kσ train The method comprises the steps of carrying out a first treatment on the surface of the The higher the upper limit of the threshold value is, or the lower limit is, the fewer the number of data points exceeding the threshold value is in the online application stage, the false alarm rate in the corresponding fault early warning result is reduced, and the false alarm rate is increased;
6) In the online application stage, based on real-time operation data point d and on DA-RNN model trained in the offline stage, an estimated residual value r of the actual measured value of the data point d minus the estimated value of the model is obtained d ；
7) Selecting a value of k, determining upper and lower thresholds, if r d Exceeding the corresponding threshold of the current k value, i.e. greater than the upper threshold limit U r (k) Or less than the threshold lower limit L r (k) Calculating the number of data points count1 (k) which continuously exceed the threshold value before the data point d, and calculating the percentage value count2 (k)% of the number of data points which exceed the threshold value in the time range of day before the data point d, if r d At the upper threshold limit U r (k) And a lower threshold limit L r (k) Outputting two judgment results 0;
8) Setting multiple discriminants, wherein the multiple discriminants are continuous overrun discriminants combined with percentage overrun discriminants, setting a continuous overrun discriminant threshold parameter S (k) and a percentage overrun discriminant threshold parameter P (k)%, recording the judgment result of the k-S (k) early warning strategy combination as 1 when the condition count1 (k) is more than or equal to S (k) is met, recording 0 when the condition count2 (k) is not met, recording 1 when the condition count2 (k) is more than or equal to P (k) is met, and recording 0 when the condition is not met;
9) Repeating the step 7) and the step 8) for all other values of k to obtain a judging result sequence, and judging whether a final alarm is given for the real-time data point d based on the judging result sequence.
Further, in the step 3), the encoder and the partial flow of the input attention mechanism are as shown in fig. 4, and the input is a sliding window related variableWherein n is the number of related variables, the input attention mechanism part inputs the related variables X, the hidden state output h and the memory unit output s of the encoder, and the influence weight of the kth related variable on the target variable is output->The calculation process is as follows:
wherein v is en ,W en ,U en For the parameter to be learned, at time t w Based on impact weightsReconstructing the related variable to obtain a time t w Is expressed as:
the reconstructed vector is input as an encoder part, the encoder is composed of a plurality of LSTM units, and the reconstructed vector is output as a hidden state h and a memory unit s, at a time t w The update of LSTM is performed by forget gate f t Input gate i t Output gate o t The decision, its update rule is as follows:
wherein W is f ,W i ,W o ,W s ,b f ,b i ,b o ,b s For the parameter to be learned, σ is a sigmoid function, and the corresponding elements are multiplied by the ";
the partial flow of the decoder and the time attention mechanism is shown in FIG. 5, the input of the time attention mechanism is the hidden state h, the output h 'of the hidden state of the decoder and the output s' of the memory unit, and the weight of the ith hidden state is outputThe calculation process is as follows:
wherein v is de ,W de ,U de Reconstructing the hidden state based on the weight beta to obtain a semantic vector for the parameter to be learned, wherein the semantic vector is expressed as follows:
then, integrating the semantic vector and the target variable historical value by using a linear regression model to obtain the input of the decoderExpressed as:
wherein the method comprises the steps ofIs a parameter to be learned;
the decoder is also composed of LSTM units, forgetting the gate f t ' input gate i t ' output door o t The update rule of' is as follows:
wherein W' f ,W′ i ,W′ o ,W′ s ,b′ f ,b′ i ,b′ o ,b′ s Is a parameter to be learned;
finally, obtaining the estimated value of the T target variable at the current moment through a linear functionThe following is shown:
wherein W is T ,b T Is a parameter to be learned.
An example of the present application is given below, and specific steps performed by this example are described in detail in connection with tables 1, 2, and fig. 2-8.
According to the method, fault early warning is conducted on a wind turbine generator set, a non-drive end bearing assembly of the wind turbine generator is damaged, the non-drive end bearing of the wind turbine generator is damaged in 2017.04.17, data collected in 2016-2017 of a SCADA (supervisory control and data acquisition) system of the wind turbine generator is selected to conduct fault early warning, wherein the data sampling interval of the SCADA system is 5min, the data information is 16 months, the time range is 2016.01.01 00:00:00-2017.04.30 23:55:00, the temperature measured by a temperature measuring point of the non-drive end bearing of the wind turbine generator is selected to be a target variable, and other operating parameters of the wind turbine generator and all parameters affecting the target variable value such as system parameters are selected to be related variables. The dataset specific variables are shown in table 1:
TABLE 1 target variable and related variable for a wind farm
The implementation data set of the failure early warning method for the non-driving end bearing assembly of the wind turbine generator is the operation data of the wind turbine generator for 16 months, and the implementation steps of the method are as follows:
1) Acquiring an operation data set recorded in the SCADA system of the fan in a normal operation state, wherein the data set comprises the temperature of a non-driving end bearing of a target variable generator and all related variables, and selecting data in the normal operation state for the first 12 months, namely 2016.01.01 00:00:00 to 2016.12.31 23:55:00, to construct an initial training setData construction initial test set for the last 4 months, 2017.01.01 00:00:00 to 2017.04.30 23:55:00
2) For initial training setThe preprocessing step comprises the steps of eliminating and interpolating isolated abnormal points, eliminating and interpolating data during fan shutdown maintenance based on operation and maintenance records, and interpolating missing values, wherein in the embodiment, the cut-in wind speed is 2m/s, the rated wind speed is 14m/s, the cut-out wind speed is 25m/s, the rated power is 1500kW, and the test set->As a real-time operation data set in online application, training set and test set data before target variable generator non-driving end bearing temperature preprocessing are shown in fig. 2, missing values and data during shutdown are both shown in 0 value, and preprocessed training set and test set data are shown in fig. 3;
3) Selecting a DA-RNN model as a variable estimation modelSelecting the length W=5 of a sliding window, wherein the moment in the sliding window is t w W=t-4, T-3,.. assigning each time T by designing an input attention mechanism w Related variableFor the target variable->Influence weight of->Reconstruction related variable +.>As encoder input, the encoder part is a plurality of LSTM units, the input of each LSTM unit is a reconstruction related variable at one moment in the sliding window, and the encoder part outputs as a hidden vector h; then designing a time attention mechanism to distribute hidden vectors of different historical moments in the sliding window to target variables of the current moment +.>Influence weight of->Obtaining a semantic vector c, integrating the semantic vector and a target variable history value by using a linear regression model as input of a decoder, wherein the decoder is divided into a plurality of LSTM units and outputs the LSTM units as a time t before the current time T-1 Obtaining the estimated value of the target variable at the current moment through a linear function>After training a model through a training set, inputting a test set to obtain a real-time estimated value of the test set, wherein a variable estimated result is shown in fig. 6, and a black circle is an actual running value of a target variable corresponding to a fault confirmation moment;
4) Subtracting the model output at the corresponding moment from the actual value of the target variable in the training set, namely the target variableObtaining a training set estimated residual sequence by the quantity estimation value, and obtaining a residual sequence average value mu train Standard deviation sigma train ；
5) Multi-threshold setting is carried out based on the residual sequence of the training set, and the multi-threshold setting is that the average value mu of the estimated residual sequence of the training set is set train Plus or minus k times standard deviation sigma train Respectively used as the upper and lower limits of the residual sequence threshold, wherein the upper limit U r (k)＝μ train +kσ train Lower limit L r (k)＝μ train -kσ train Wherein k has a value of [1.5,2,2.5,3]]In fig. 7, a black solid line is a plurality of upper threshold limits, and a black broken line is a plurality of lower threshold limits;
6) The test set is used as an online application stage data set, based on real-time operation data point d in the test set, the DA-RNN model which is trained through an offline stage is used for obtaining an estimated residual value r of the actual measured value of the data point d minus the estimated value of the model d ；
7) Selecting a value of k, determining upper and lower thresholds, if r d Exceeding the corresponding threshold of the current k value, i.e. greater than the upper threshold limit U r (k) Or less than the threshold lower limit L r (k) Calculating the number of data points count1 (k) which continuously exceed the threshold value before the data point d, and calculating the percentage value count2 (k)% of the number of data points which exceed the threshold value in the time range of day before the data point d, if r d At the upper threshold limit U r (k) And a lower threshold limit L r (k) Outputting two judgment results 0;
8) Setting multiple discriminant criteria, wherein the multiple discriminant criteria are continuous overrun discriminant criteria combined with percentage overrun discriminant criteria, setting continuous overrun discriminant criteria threshold parameter S (k) and percentage overrun discriminant criteria threshold parameter P (k)%, and setting the parameters as in the embodimentWhen the condition count1 (k) is not less than S (k), the judgment result of the k-S (k) early warning strategy combination is marked as 1, when the condition is not met, the judgment result is marked as 0, and when the condition count2 (k) is not less than P (k), the judgment result of the k-P (k) early warning strategy combination is marked as 1If 1 is recorded, if the condition is not satisfied, 0 is recorded as the judgment result;
9) Repeating the steps 7) and 8) for all other values of k, outputting two 0-1 judgment results for each k value, and finally outputting a 0-1 sequence for the data point d, wherein the sequence dimension is 8 dimensions in the embodiment, when the number of 1 s in the sequence is greater than 4, the data point d is given an alarm, when the number of 1 s in the sequence is less than or equal to 4, no alarm is given, the operation is executed for all the data points of the test set, the final early warning result is shown in the table, the black star is the data point for giving the alarm, the black circle is the data point at the fault confirmation time, the earliest alarm time is 2017.03.23 02:00:00, and the judgment result at the time point is shown in the table below:
table 2 determination of earliest alarm time
| Parameter selection | Percentage overrun | Continuous overrun |
| k＝1.5,S＝P＝60 | 0 | 1 |
| k＝2,S＝P＝45 | 1 | 1 |
| k＝2.5,S＝P＝36 | 1 | 1 |
| k＝3,S＝P＝30 | 0 | 1 |
The judging result sequence is [0,1,1,1,1,1,0,1], and the number of 1 is 6, so that an alarm is finally given at the moment, and early warning is realized 24 days before the occurrence of the fault.
The invention discloses a fault early warning method for key components of a wind turbine generator, which mainly comprises links of target variable and related variable selection, data preprocessing, variable estimation model training, real-time operation residual error acquisition, robustness early warning strategy design and the like. Fig. 1 is a specific flow chart of fault early warning of a key component of the present invention, fig. 2, fig. 3 is a data chart before and after preprocessing a selected target variable in an embodiment of the present invention, fig. 4 is a schematic diagram of an encoder for adding an input attention mechanism to a selected DA-RNN model of the present invention, fig. 5 is a schematic diagram of a decoder for adding a time attention mechanism to a selected DA-RNN model of the present invention, fig. 6 is a target variable estimation result chart of a variable estimation model of an embodiment of the present invention, fig. 7 is a multi-threshold setting schematic diagram of an embodiment of the present invention, and fig. 8 is a early warning result chart of an embodiment of the present invention, which shows that the present invention can realize accurate warning before occurrence of a fault, and the result has effectiveness and reliability.
The above-described embodiments are merely examples of the present invention, and although the best examples of the present invention and the accompanying drawings are disclosed for illustrative purposes, it will be understood by those skilled in the art that: various alternatives, variations and modifications are possible without departing from the spirit and scope of the invention and the appended claims. Therefore, the present invention should not be limited to the preferred embodiments and the disclosure of the drawings.
Claims (6)
1. A fault early warning method for key components of a wind turbine generator based on DA-RNN is characterized by comprising the following steps:
1) Selecting a wind turbine generator to be subjected to fault early warning, acquiring N pieces of operation data recorded in a SCADA (supervisory control and data acquisition) system of the fan in a normal operation state, and selecting a temperature variable measured at a temperature measuring point of a component to be early warned in the SCADA system as a target variable y, wherein the temperature variable is related to the temperature of the componentWith variables as related variables X, constructing an initial training set
2) Offline training stage, based on initial training setDesigning a data preprocessing flow, wherein the preprocessing step comprises the steps of eliminating and interpolating isolated abnormal points, eliminating and interpolating data during fan shutdown maintenance based on operation and maintenance records, interpolating missing values, and preprocessing a training set [ X ] train ,y train ]Performing model training as input of a variable estimation model;
3) Selecting a circulating neural network DA-RNN model based on a double-attention mechanism as a variable estimation model, selecting a sliding window length W, and expressing the moment in the sliding window as t w W=t-w+1, T-w+2, &..t, where T is the current time, each time T is assigned by designing an input attention mechanism w Related variableFor the target variable->Influence weight of->Reconstruction related variable +.>As encoder input, the encoder part is a plurality of LSTM units, the input of each LSTM unit is a reconstruction related variable at one moment in the sliding window, and the encoder part outputs as a hidden vector h; then designing a time attention mechanism to distribute hidden vectors of different historical moments in the sliding window to target variables of the current moment +.>Influence weight of->Obtaining a semantic vector c, integrating the semantic vector and a target variable history value by using a linear regression model as input of a decoder, wherein the decoder is divided into a plurality of LSTM units and outputs the LSTM units as a time t before the current time T-1 Obtaining the estimated value of the target variable at the current moment through a linear function>
4) Subtracting the model output at the corresponding moment from the actual value of the target variable in the training set, namely the estimated value of the target variable, obtaining an estimated residual sequence of the training set, and obtaining the average value mu of the residual sequence train Standard deviation sigma train ；
5) Multi-threshold setting is carried out based on the residual sequence of the training set, and the multi-threshold setting is that the average value mu of the estimated residual sequence of the training set is set train Plus or minus k times standard deviation sigma train Respectively used as the upper and lower limits of the residual sequence threshold, wherein the upper limit U r (k)＝μ train +kσ train Lower limit L r (k)＝μ train -kσ train ；
6) In the online application stage, based on real-time operation data point d and on DA-RNN model trained in the offline stage, an estimated residual value r of the actual measured value of the data point d minus the estimated value of the model is obtained d ；
7) Selecting a value of k, determining upper and lower thresholds, if r d Exceeding the corresponding threshold of the current k value, i.e. greater than the upper threshold limit U r (k) Or less than the threshold lower limit L r (k) Calculating the number of data points count1 (k) which continuously exceed the threshold value before the data point d, and calculating the percentage value count2 (k)% of the number of data points which exceed the threshold value in the time range of day before the data point d, if r d At the upper threshold limit U r (k) And a lower threshold limit L r (k) Outputting two judgment results 0;
8) Setting multiple discriminants, wherein the multiple discriminants are continuous overrun discriminants combined with percentage overrun discriminants, setting a continuous overrun discriminant threshold parameter S (k) and a percentage overrun discriminant threshold parameter P (k)%, recording the judgment result of the k-S (k) early warning strategy combination as 1 when the condition count1 (k) is more than or equal to S (k) is met, recording 0 when the condition count2 (k) is not met, recording 1 when the condition count2 (k) is more than or equal to P (k) is met, and recording 0 when the condition is not met;
9) Repeating the step 7) and the step 8) for all other values of k to obtain a judging result sequence, and judging whether a final alarm is given for the real-time data point d based on the judging result sequence.
2. The method for early warning of faults of key components of a wind turbine generator system based on DA-RNN according to claim 1, wherein in the step 2), the pretreatment flow of offline training comprises the following steps:
a) The isolated abnormal point is usually a recording error caused by sensor abnormality, and is judged by an operation mechanism, and the judgment conditions are as follows: for temperature variables, the value is greater than 150 degrees or less than 0 degrees; for the wind speed variable, the numerical value is larger than the cut-out wind speed of the unit or smaller than the cut-in wind speed; for the power variable, the numerical value is larger than the rated power of the unit or is a negative value; when the conditions are met, the data are judged to be isolated abnormal points and are removed;
b) When the wind turbine generator is in an operation, maintenance and overhaul period, the wind turbine generator is in a shutdown state, and the data recorded by the SCADA system is usually 0 value or a system default value, wherein the value cannot represent the normal operation state of the wind turbine generator, so that the data in the shutdown, maintenance and overhaul period are removed;
c) In order to ensure the time continuity of the data, the missing values in the removed data and the SCADA system are interpolated, wherein the interpolation method is mean value interpolation, namely, the mean value of the data 1 hour before the variable interpolation position is taken as the interpolation value at the current moment.
3. The method for early warning of faults of key components of a wind turbine generator based on DA-RNN as claimed in claim 1, wherein in the step 3), the model input is related to a sliding windowVariable(s)Wherein n is the number of related variables, the input attention mechanism part inputs the related variables X, the hidden state output h and the memory unit output s of the encoder, and the influence weight of the kth related variable on the target variable is output->The calculation process is as follows:
wherein v is en ,W en ,U en For the parameter to be learned, at time t w Based on impact weightsReconstructing the related variable to obtain a time t w Is expressed as:
the reconstructed vector is input as an encoder part, the encoder is composed of a plurality of LSTM units, and the reconstructed vector is output as a hidden state h and a memory unit s, at a time t w The update of LSTM is performed by forget gate f t Input gate i t Output gate o t The decision, its update rule is as follows:
wherein W is f ,W i ,W o ,W s ,b f ,b i ,b o ,b s For the parameter to be learned, σ is a sigmoid function, and the corresponding elements are multiplied by the ";
the input of the time attention mechanism is a hidden state h, a decoder hidden state output h 'and a memory unit output s', and the weight of the ith hidden state is outputThe calculation process is as follows:
wherein v is de ,W de ,U de Reconstructing the hidden state based on the weight beta to obtain a semantic vector for the parameter to be learned, wherein the semantic vector is expressed as follows:
after which the wire is usedIntegrating semantic vectors and historical values of target variables by using a sexual regression model to obtain input of a decoderExpressed as:
wherein the method comprises the steps ofIs a parameter to be learned;
the decoder is also composed of LSTM units, forgetting the gate f t ' input gate i t ' output door o t The update rule of' is as follows:
wherein W' f ,W′ i ,W′ o ,W′ s ,b′ f ,b′ i ,b′ o ,b′ s Is a parameter to be learned;
finally, obtaining the current time T through a linear functionEstimated value of target variableThe following is shown:
wherein W is T ,b T Is a parameter to be learned.
4. The method for early warning faults of key components of a wind turbine generator system based on DA-RNN according to claim 1, wherein in the step 5), the training set residual sequence is assumed to be subjected to normal distribution, and k in the multi-threshold setting is [1.5,2,2.5,3] based on normal distribution characteristics.
5. The method for early warning faults of key components of a wind turbine generator system based on DA-RNN as claimed in claim 1, wherein in the step 8), the multiple discriminant criteria correspond to different abnormal characteristics, the continuous overrun discriminant criteria correspond to abnormal characteristics with continuous abnormal values, the percentage overrun discriminant criteria correspond to abnormal conditions with large data fluctuation, and the continuous overrun discriminant criteria threshold parameter S k Threshold parameter P of percentage overrun criterion k % is set asWherein C is a constant.
6. The method for early warning faults of key components of a wind turbine generator system based on DA-RNN according to claim 1 is characterized in that in the step 9), for a real-time data point d, two 0-1 judgment results are output for each k value, a 0-1 sequence is finally output, the dimension is 2k dimensions, when the number of 1 in the sequence is greater than k, an alarm is given to the real-time data point, and when the number of 1 in the sequence is less than or equal to k, no alarm is given.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010573207.3A CN111814849B (en) | 2020-06-22 | 2020-06-22 | DA-RNN-based wind turbine generator set key component fault early warning method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010573207.3A CN111814849B (en) | 2020-06-22 | 2020-06-22 | DA-RNN-based wind turbine generator set key component fault early warning method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111814849A CN111814849A (en) | 2020-10-23 |
| CN111814849B true CN111814849B (en) | 2024-02-06 |
Family
ID=72845420
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010573207.3A Active CN111814849B (en) | 2020-06-22 | 2020-06-22 | DA-RNN-based wind turbine generator set key component fault early warning method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111814849B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112581031B (en) * | 2020-12-30 | 2023-10-17 | 杭州朗阳科技有限公司 | Method for implementing real-time monitoring of motor abnormality by Recurrent Neural Network (RNN) through C language |
| CN113516273B (en) * | 2021-04-02 | 2024-06-04 | 中国船舶重工集团公司军品技术研究中心 | Diesel engine supercharger fault prediction method for power generation |
| CN114372504B (en) * | 2021-12-06 | 2024-09-10 | 燕山大学 | Wind turbine generator system fault early warning method based on graph neural network |
| CN114460481B (en) * | 2022-01-27 | 2024-09-03 | 浙江商储数智能源科技有限公司 | Bi-LSTM and attention mechanism-based energy storage battery thermal runaway early warning method |
| CN114429249B (en) * | 2022-04-06 | 2022-08-16 | 杭州未名信科科技有限公司 | Method, system, equipment and storage medium for predicting service life of steel pipe bundle production equipment |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104035431A (en) * | 2014-05-22 | 2014-09-10 | 清华大学 | Obtaining method and system for kernel function parameters applied to nonlinear process monitoring |
| CN104102773A (en) * | 2014-07-05 | 2014-10-15 | 山东鲁能软件技术有限公司 | Equipment fault warning and state monitoring method |
| CN104914851A (en) * | 2015-05-21 | 2015-09-16 | 北京航空航天大学 | Adaptive fault detection method for airplane rotation actuator driving device based on deep learning |
| WO2016145850A1 (en) * | 2015-03-19 | 2016-09-22 | 清华大学 | Construction method for deep long short-term memory recurrent neural network acoustic model based on selective attention principle |
| CN107065843A (en) * | 2017-06-09 | 2017-08-18 | 东北大学 | Multi-direction KICA batch processes fault monitoring method based on Independent subspace |
| CN107609574A (en) * | 2017-08-18 | 2018-01-19 | 上海电力学院 | Wind turbines fault early warning method based on data mining |
| CN107977508A (en) * | 2017-11-29 | 2018-05-01 | 北京优利康达科技股份有限公司 | A kind of dynamo bearing failure prediction method |
| CN109740175A (en) * | 2018-11-18 | 2019-05-10 | 浙江大学 | A kind of point judging method that peels off towards Wind turbines power curve data |
| CN110298455A (en) * | 2019-06-28 | 2019-10-01 | 西安因联信息科技有限公司 | A kind of mechanical equipment fault intelligent early-warning method based on multivariable estimation prediction |
| CN111079343A (en) * | 2019-12-04 | 2020-04-28 | 浙江大学 | Wind turbine generator effective wind speed estimation method based on width learning |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109710636B (en) * | 2018-11-13 | 2022-10-21 | 广东工业大学 | Unsupervised industrial system anomaly detection method based on deep transfer learning |
-
2020
- 2020-06-22 CN CN202010573207.3A patent/CN111814849B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104035431A (en) * | 2014-05-22 | 2014-09-10 | 清华大学 | Obtaining method and system for kernel function parameters applied to nonlinear process monitoring |
| CN104102773A (en) * | 2014-07-05 | 2014-10-15 | 山东鲁能软件技术有限公司 | Equipment fault warning and state monitoring method |
| WO2016145850A1 (en) * | 2015-03-19 | 2016-09-22 | 清华大学 | Construction method for deep long short-term memory recurrent neural network acoustic model based on selective attention principle |
| CN104914851A (en) * | 2015-05-21 | 2015-09-16 | 北京航空航天大学 | Adaptive fault detection method for airplane rotation actuator driving device based on deep learning |
| CN107065843A (en) * | 2017-06-09 | 2017-08-18 | 东北大学 | Multi-direction KICA batch processes fault monitoring method based on Independent subspace |
| CN107609574A (en) * | 2017-08-18 | 2018-01-19 | 上海电力学院 | Wind turbines fault early warning method based on data mining |
| CN107977508A (en) * | 2017-11-29 | 2018-05-01 | 北京优利康达科技股份有限公司 | A kind of dynamo bearing failure prediction method |
| CN109740175A (en) * | 2018-11-18 | 2019-05-10 | 浙江大学 | A kind of point judging method that peels off towards Wind turbines power curve data |
| CN110298455A (en) * | 2019-06-28 | 2019-10-01 | 西安因联信息科技有限公司 | A kind of mechanical equipment fault intelligent early-warning method based on multivariable estimation prediction |
| CN111079343A (en) * | 2019-12-04 | 2020-04-28 | 浙江大学 | Wind turbine generator effective wind speed estimation method based on width learning |
Non-Patent Citations (3)
| Title |
|---|
| 基于MSET的电站风机故障预警技术研究;杨婷婷;张蓓;吕游;邸小慧;热能动力工程;第32卷(第9期);全文 * |
| 非正常工况下化工过程设备故障实时关联预警研究;胡瑾秋;张来斌;伊岩;蔡爽;中国安全科学学报;第26卷(第9期);全文 * |
| 风电机组齿轮箱系统的故障预警研究;黄小光;潘东浩;史晓鸣;王杏;王欣;;机电工程(第06期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111814849A (en) | 2020-10-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111814849B (en) | DA-RNN-based wind turbine generator set key component fault early warning method | |
| Udo et al. | Data-driven predictive maintenance of wind turbine based on SCADA data | |
| CN111597682B (en) | Method for predicting remaining life of bearing of gearbox of wind turbine | |
| CN111537219B (en) | Fan gearbox performance detection and health assessment method based on temperature parameters | |
| CN111539553B (en) | Wind turbine generator fault early warning method based on SVR algorithm and off-peak degree | |
| Yang et al. | Fault detection of wind turbine generator bearing using attention-based neural networks and voting-based strategy | |
| CN104952000A (en) | Wind turbine operating state fuzzy synthetic evaluation method based on Markov chain | |
| CN111415070A (en) | Wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data | |
| CN111814848B (en) | Self-adaptive early warning strategy design method for temperature faults of wind turbine generator | |
| CN114215706B (en) | Early warning method and device for cracking faults of wind turbine generator blades | |
| CN112132394B (en) | Power plant circulating water pump predictive state evaluation method and system | |
| CN117195121A (en) | Wind turbine generator abnormal state identification method and system based on improved countermeasure automatic encoder | |
| CN115578084A (en) | Wind turbine generator set frequency converter fault early warning method based on deep convolution self-encoder | |
| CN114048767A (en) | Fault monitoring and early warning method for wind power master control system | |
| Zheng et al. | Rotating machinery fault prediction method based on Bi-LSTM and attention mechanism | |
| Zhu et al. | Power data preprocessing method of mountain wind farm based on POT-DBSCAN | |
| CN116793666A (en) | Wind turbine generator system gearbox fault diagnosis method based on LSTM-MLP-LSGAN model | |
| CN115935805A (en) | Wind power gearbox bearing health state assessment method and system based on machine learning | |
| Song et al. | Anomaly detection of wind turbine generator based on temporal information | |
| CN115577854A (en) | Quantile regression wind speed interval prediction method based on EEMD-RBF combination | |
| Souza et al. | Evaluation of data based normal behavior models for fault detection in wind turbines | |
| Qi et al. | Fault diagnosis in wind turbines based on weighted joint domain adversarial network under various working conditions | |
| Wang et al. | Remaining Life Prediction for High-speed Rail Bearing Considering Hybrid Data-model-driven Approach | |
| Zhenhao et al. | Prediction of wind power ramp events based on deep neural network | |
| Dhiman et al. | Enhancing wind turbine reliability through proactive high speed bearing prognosis based on adaptive threshold and gated recurrent unit networks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |