CN111814849A - DA-RNN-based wind turbine generator key component fault early warning method - Google Patents
DA-RNN-based wind turbine generator key component fault early warning method Download PDFInfo
- Publication number
- CN111814849A CN111814849A CN202010573207.3A CN202010573207A CN111814849A CN 111814849 A CN111814849 A CN 111814849A CN 202010573207 A CN202010573207 A CN 202010573207A CN 111814849 A CN111814849 A CN 111814849A
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000007781 pre-processing Methods 0.000 claims abstract description 23
- 238000012423 maintenance Methods 0.000 claims abstract description 21
- 238000013528 artificial neural network Methods 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims description 43
- 239000013598 vector Substances 0.000 claims description 28
- 230000007246 mechanism Effects 0.000 claims description 19
- 230000002159 abnormal effect Effects 0.000 claims description 17
- 230000000875 corresponding effect Effects 0.000 claims description 15
- 230000008030 elimination Effects 0.000 claims description 8
- 238000003379 elimination reaction Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 238000012886 linear function Methods 0.000 claims description 7
- 238000012417 linear regression Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 230000002596 correlated effect Effects 0.000 claims description 4
- 238000005259 measurement Methods 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 3
- 230000005856 abnormality Effects 0.000 claims description 2
- 238000013461 design Methods 0.000 abstract description 6
- 125000004122 cyclic group Chemical group 0.000 abstract description 3
- 238000013480 data collection Methods 0.000 abstract 1
- 238000012544 monitoring process Methods 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 12
- 238000012360 testing method Methods 0.000 description 8
- 230000008901 benefit Effects 0.000 description 3
- 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
- 239000007787 solid Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
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 wind turbine generator key component fault early warning method based on a double-attention-machine system and a cyclic neural network DA-RNN. The method comprises the steps of designing a preprocessing flow and selecting a DA-RNN model to carry out variable real-time estimation based on a data collection and monitoring control (SCADA) system data set collected in a normal operation state of the wind turbine generator, outputting a judgment result sequence through multi-threshold setting and judgment criterion design, and giving a final early warning result according to the judgment result sequence. In the fault early warning method, the preprocessing flow is designed aiming at different types of noise data, so that a reliable data basis 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 guaranteed; the multi-threshold setting and judgment criterion design avoids single 0-1 judgment, so that the final early warning result has more robustness; finally, the fault early warning of key components is realized, the unit downtime is reduced, the operation and maintenance cost is saved, and the theoretical performance and the practicability are high.
Description
Technical Field
The invention relates to a DA-RNN-based wind turbine generator system key component fault early warning method, which is a method for early warning key component faults 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 estimate a target variable in real time, and designing a fault early warning strategy selected by combining multi-threshold setting and multi-discriminant criteria based on real-time running residual errors.
Background
With global pollution and increasingly scarcity of traditional fossil energy, development of clean energy has attracted extensive attention, wind energy is rapidly developed with the advantage of cleanness and no pollution, and the wind power industry is one of novel renewable energy industries which are vigorously developed at home and abroad. At present, the total installed capacity of a fan in China is located at the front of the world, but in recent years, the preparation in the research and development period is insufficient due to the rapid development of the wind power generation market, 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 multiple components and multiple subsystems, the wind turbine generator generally operates in remote areas such as suburb plains, mountainous areas, near sea and the like, the operating environment is severe and changeable, and the failure of a key component can cause shutdown maintenance of the whole machine, so that a large amount of economic loss is brought. Therefore, the early identification of the abnormity of the key component is realized, the phenomenon that the abnormity of the initial stage is changed into a catastrophic fault is avoided, and the fault early warning of the key component is realized, so that the predictive maintenance is carried out, and the method has great significance for 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 accuracy of an early warning result is difficult to ensure by the existing simple early warning strategy. Therefore, a more accurate variable estimation model is selected, and a robust early warning strategy is designed, so that the method has great significance for realizing accurate fault early warning.
Disclosure of Invention
The invention aims to carry out fault early warning on a key component of a wind turbine generator by accurately estimating a target variable and designing a robustness early warning strategy, and provides a DA-RNN-based fault early warning method for the key component of the wind turbine generator. According to the method, a data set of a normal operation state of the wind turbine generator is selected, a data preprocessing flow is designed by considering different noise data types, a DA-RNN is selected as a variable estimation model, the influence of related variables and historical information is comprehensively considered to estimate a target variable in real time, the estimation accuracy is guaranteed, an early warning strategy of robustness 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 expanded to all key components of the wind turbine generator with temperature measuring points, achieves fault early warning of the key components, and has practical value and strong expansibility.
The purpose of the invention is realized by the following technical scheme: a DA-RNN-based wind turbine generator key component fault early warning method comprises 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 fan SCADA system under a normal operation state, wherein key components of the wind turbine generator comprise a gear box, a generator, a pitch control system and the like, selecting a temperature variable measured by a temperature measuring point of a component to be subjected to early warning in the SCADA system as a target variable y, taking all variables related to the temperature of the component as related variables X, and constructing an initial training set
2) An off-line training phase based on an initial training setDesigning a data preprocessing flow, wherein the preprocessing step comprises the elimination and interpolation of isolated abnormal points, the elimination and interpolation of data during the shutdown maintenance of the fan based on operation and maintenance records, the interpolation of missing values, and the preprocessing of a training set [ X ]train,ytrain]Performing model training as input of a variable estimation model;
3) selecting a cyclic neural network DA-RN based on a double-attention machine systemN model is used as a variable estimation model, the length W of a sliding window is selected, and the time in the sliding window is represented as twT, where T is the current time, each time T is assigned by the design input attention mechanismwDependent variableFor target variableInfluence weight ofReconstructing correlated variablesThe encoder part is used as the input of an encoder, the encoder part is a plurality of LSTM units, the input of each LSTM unit is a reconstruction related variable of a moment in a sliding window, and the output of the encoder part is a hidden vector h; designing time attention mechanism to distribute hidden vectors at different historical moments in sliding window to target variables at current momentInfluence weight ofObtaining a semantic vector c, integrating the semantic vector and a target variable historical value by using a linear regression model as the input of a decoder, wherein the decoder is partially composed of a plurality of LSTM units, and the output is the last moment t of the current momentT-1The estimated value of the target variable at the current moment is obtained by the 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, to obtain the estimated residual sequence of the training set, and calculating the mean value mu of the residual sequencetrainAnd standard deviation sigmatrain;
5) Based on trainingPerforming multi-threshold setting on the training set residual sequence, wherein the multi-threshold setting is the mean value mu of the training set estimated residual sequencetrainPlus or minus k times standard deviation sigmatrainRespectively as the upper and lower limits of the threshold of the residual sequence, wherein the upper limit Ur(k)=μtrain+kσtrainLower limit of Lr(k)=μtrain-kσtrain(ii) a 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 in the online application stage is, the lower the false alarm rate in the corresponding fault early warning result is, and the higher the false alarm rate is;
6) in the on-line application stage, based on the real-time operation data point d and the DA-RNN model trained in the off-line stage, the estimation residual value r of the model estimation value subtracted from the actual measurement value of the data point d is obtainedd;
7) Selecting a value of k, determining the upper and lower limits of the threshold value, if rdExceeds the corresponding threshold value of the current k value, namely is greater than the upper limit U of the threshold valuer(k) Or less than the lower threshold limit Lr(k) Calculating the number count1(k) of data points which are continuously over the threshold value before the data point d, and the percentage value count2 (k)% of the number of data points which are over the threshold value in the time range of one day before the data point d in the total number of data points in one day if rdAt the upper threshold Ur(k) And a lower threshold Lr(k) Meanwhile, two judgment results 0 are output;
8) setting multiple judgment criteria, wherein the multiple judgment criteria are continuous overrun judgment criteria combined with percentage overrun judgment criteria, setting a threshold parameter S (k) of the continuous overrun judgment criteria and a threshold parameter P (k)%) of the percentage overrun judgment criteria under a current k value, recording a judgment result of a k-S (k) early warning strategy combination as 1 when a condition count1(k) is greater than or equal to S (k), recording a judgment result as 0 when the condition is not met, recording a judgment result of a k-P (k) early warning strategy combination as 1 when a condition count2(k) is greater than or equal to P (k), and recording a judgment result as 0 when the condition is not met;
9) and repeating the step 7) and the step 8) for all other values of k to obtain a discrimination result sequence, and judging whether to give final alarm to the real-time data point d or not based on the discrimination result sequence.
Further, in the step 2), the preprocessing procedure of the offline training includes the following steps:
a) the isolated abnormal point is usually a recording error caused by the abnormality of the sensor, and is judged through an operation mechanism, wherein the judgment condition is as follows: for temperature variations, the value is greater than 150 degrees or less than 0 degrees; for the wind speed variable, the value is larger than the cut-in wind speed of the unit or smaller than the cut-in wind speed; for the power variable, the value is larger than the rated power of the unit or is a negative value; when the above conditions are met, the data is judged to be isolated abnormal points and is removed;
b) checking by combining with operation and maintenance records, when the wind turbine generator is in an operation, maintenance and overhaul period, the wind turbine generator is in a shutdown state, data recorded by an SCADA system is usually a 0 value or a system default value, and the value cannot represent a normal operation state of the wind turbine generator, so that data in the shutdown, maintenance and overhaul period are removed;
c) in order to ensure the time continuity of the data, the removed data and the missing value in 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 input is a relevant variable in a sliding windowWherein n is the number of the relevant variables, the input attention mechanism part inputs the relevant variables X, the hidden state output h and the memory unit output s of the encoder, and outputs the influence weight of the kth relevant variable on the target variableThe calculation process is as follows:
wherein v isen,Wen,UenFor learning as requiredAt time t, ofwBased on impact weightReconstructing the related variable to obtain the time twIs represented as:
the reconstructed vector is input as part of an encoder, which is composed of a plurality of LSTM units and outputs a hidden state h and a memory unit s at a time twUpdate of LSTM by forgetting gate ftInput door itOutput gate otThe decision, the update rule is as follows:
wherein Wf,Wi,Wo,Ws,bf,bi,bo,bsFor the parameter to be learned, σ is a sigmoid function, which represents multiplication of corresponding elements;
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 isde,Wde,UdeReconstructing the hidden state for the parameter needing to be learned based on the weight beta to obtain a semantic vector, wherein the semantic vector is expressed as:
then, the linear regression model is used for integrating the semantic vector and the target variable historical value to obtain the input of the decoderExpressed as:
the decoder is also composed of LSTM units, forgetting gate ft', input door it', output gate otThe update rule of' is as follows:
wherein W'f,W′i,W′o,W′s,b′f,b′i,b′o,b′sParameters required to be learned;
finally, obtaining the estimated value of the target variable T at the current moment through a linear functionAs follows:
wherein WT,bTIs the parameter required to be learned.
Further, in the step 5), it may be assumed that the training set residual sequence follows normal distribution, and based on a normal distribution characteristic, a value of k in the multi-threshold setting is [1.5,2,2.5,3 ].
Further, in the step 8), the multiple criteria correspond to different abnormal characteristics, the continuous out-of-limit criteria correspond to abnormal characteristics with continuous abnormal values, the percentage out-of-limit criteria correspond to abnormal conditions with large data fluctuation, and the threshold parameter S of the continuous out-of-limit criteriakAnd a percentage overrun criterion threshold parameter Pk% set toWherein C is a constant.
Further, in step 9), for the real-time data point d, two 0-1 judgment results are output for each k value, and finally a 0-1 sequence with 2k dimensions is output, when the number of 1 s in the sequence is greater than k, an alarm is given to the real-time data point, and when the number of 1 s 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) designing a normalized data preprocessing flow aiming at different types of noise data existing in a data set in a normal operation period, ensuring that a constructed training set can accurately represent the normal operation state of a wind turbine generator, and providing a data basis for accurate estimation of a target variable;
2) the DA-RNN is selected as a variable estimation model, the influence of related variables and historical information is comprehensively considered, different influence weights of the DA-RNN on target variable estimation are determined through a double-layer attention mechanism, and the reliability and the accuracy of the model are guaranteed;
3) the method comprises the steps of designing a robust fault early warning strategy, comprehensively considering the requirements of early warning results on different false alarm rates and missed report rates by multi-threshold setting, carrying out more comprehensive early warning on different abnormal characteristics by two discrimination criteria, giving out a final early warning result based on a judgment result sequence, and comprehensively considering different early warning strategy combinations, 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, and the process is applicable to all critical components of a wind turbine generator with corresponding temperature measuring points and has expansibility.
Drawings
FIG. 1 is a flow chart of a wind turbine generator key component fault early warning method of the present invention;
FIG. 2 is a data diagram of a selected target variable prior to preprocessing in an embodiment of the present invention;
FIG. 3 is a graph of data after preprocessing of selected target variables in an embodiment of the present invention;
FIG. 4 is a schematic diagram of an encoder incorporating the DA-RNN model of the present invention with 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 diagram of a target variable estimation result of the variable estimation model according to the embodiment of the present invention;
FIG. 7 is a schematic diagram of a multiple threshold setting of an embodiment of the present invention;
fig. 8 is a diagram of an early warning result according to an embodiment of the present invention.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail below.
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 than those specifically described and will be readily apparent to those of ordinary skill in the art without departing from the spirit of the present invention, and therefore the present invention is not limited to the specific embodiments disclosed below.
As shown in fig. 1, the method for early warning of the fault of the key component of the wind turbine generator based on the DA-RNN provided by the present application includes:
1) selecting a wind turbine generator to be subjected to fault early warning, acquiring N pieces of operation data recorded in a fan SCADA system under a normal operation state, wherein key components of the wind turbine generator comprise a gear box, a generator, a pitch control system and the like, selecting a temperature variable measured by a temperature measuring point of a component to be subjected to early warning in the SCADA system as a target variable y, taking all variables related to the temperature of the component as related variables X, and constructing an initial training set
2) An off-line training phase based on an initial training setDesigning a data preprocessing flow, wherein the preprocessing step comprises the elimination and interpolation of isolated abnormal points, the elimination and interpolation of data during the shutdown maintenance of the fan based on operation and maintenance records, the interpolation of missing values, and the preprocessing of a training set [ X ]train,ytrain]Performing model training as input of a variable estimation model;
3) selection radicalSelecting a sliding window length W in a DA-RNN model of a cyclic neural network of a double-attention machine system as a variable estimation model, wherein the time in the sliding window is represented as twT, where T is the current time, each time T is assigned by the design input attention mechanismwDependent variableFor target variableInfluence weight ofReconstructing correlated variablesThe encoder part is used as the input of an encoder, the encoder part is a plurality of LSTM units, the input of each LSTM unit is a reconstruction related variable of a moment in a sliding window, and the output of the encoder part is a hidden vector h; designing time attention mechanism to distribute hidden vectors at different historical moments in sliding window to target variables at current momentInfluence weight ofObtaining a semantic vector c, integrating the semantic vector and a target variable historical value by using a linear regression model as the input of a decoder, wherein the decoder is partially composed of a plurality of LSTM units, and the output is the last moment t of the current momentT-1The estimated value of the target variable at the current moment is obtained by the 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, to obtain the estimated residual sequence of the training set, and calculating the mean value mu of the residual sequencetrainAnd standard deviation sigmatrain;
5) Performing multi-threshold setting based on the residual sequence of the training set, wherein the multi-threshold setting is the mean value mu of the estimated residual sequence of the training settrainPlus or minus k times standard deviation sigmatrainRespectively as the upper and lower limits of the threshold of the residual sequence, wherein the upper limit Ur(k)=μtrain+kσtrainLower limit of Lr(k)=μtrain-kσtrain(ii) a 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 in the online application stage is, the lower the false alarm rate in the corresponding fault early warning result is, and the higher the false alarm rate is;
6) in the on-line application stage, based on the real-time operation data point d and the DA-RNN model trained in the off-line stage, the estimation residual value r of the model estimation value subtracted from the actual measurement value of the data point d is obtainedd;
7) Selecting a value of k, determining the upper and lower limits of the threshold value, if rdExceeds the corresponding threshold value of the current k value, namely is greater than the upper limit U of the threshold valuer(k) Or less than the lower threshold limit Lr(k) Calculating the number count1(k) of data points which are continuously over the threshold value before the data point d, and the percentage value count2 (k)% of the number of data points which are over the threshold value in the time range of one day before the data point d in the total number of data points in one day if rdAt the upper threshold Ur(k) And a lower threshold Lr(k) Meanwhile, two judgment results 0 are output;
8) setting multiple judgment criteria, wherein the multiple judgment criteria are continuous overrun judgment criteria combined with percentage overrun judgment criteria, setting a threshold parameter S (k) of the continuous overrun judgment criteria and a threshold parameter P (k)%) of the percentage overrun judgment criteria under a current k value, recording a judgment result of a k-S (k) early warning strategy combination as 1 when a condition count1(k) is greater than or equal to S (k), recording a judgment result as 0 when the condition is not met, recording a judgment result of a k-P (k) early warning strategy combination as 1 when a condition count2(k) is greater than or equal to P (k), and recording a judgment result as 0 when the condition is not met;
9) and repeating the step 7) and the step 8) for all other values of k to obtain a discrimination result sequence, and judging whether to give final alarm to the real-time data point d or not based on the discrimination result sequence.
Further, in the step 3), the encoder and the input attention mechanism are partially processed as shown in fig. 4, and the input is the related variable in the sliding windowWherein n is the number of the relevant variables, the input attention mechanism part inputs the relevant variables X, the hidden state output h and the memory unit output s of the encoder, and outputs the influence weight of the kth relevant variable on the target variableThe calculation process is as follows:
wherein v isen,Wen,UenFor the parameter to be learned, at time twBased on impact weightReconstructing the related variable to obtain the time twIs represented as:
the reconstructed vector is input as part of an encoder, which is composed of a plurality of LSTM units and outputs a hidden state h and a memory unit s at a time twUpdate of LSTM by forgetting gate ftInput door itOutput gate otThe decision, the update rule is as follows:
wherein Wf,Wi,Wo,Ws,bf,bi,bo,bsFor the parameter to be learned, σ is a sigmoid function, which represents multiplication of corresponding elements;
the decoder and time attention mechanism are partially shown in FIG. 5, the input of the time attention mechanism is hidden state h, the output h 'of the hidden state of the decoder and the output s' of the memory unit, the weight of the ith hidden state is outputThe calculation process is as follows:
wherein v isde,Wde,UdeReconstructing the hidden state for the parameter needing to be learned based on the weight beta to obtain a semantic vector, wherein the semantic vector is expressed as:
then, the linear regression model is used for integrating the semantic vector and the target variable historical value to obtain decodingInput of the deviceExpressed as:
the decoder is also composed of LSTM units, forgetting gate ft', input door it', output gate otThe update rule of' is as follows:
wherein W'f,W′i,W′o,W′s,b′f,b′i,b′o,b′sParameters required to be learned;
finally, obtaining the estimated value of the target variable T at the current moment through a linear functionAs follows:
wherein WT,bTIs the parameter required to be learned.
An embodiment of the present application is given below, and specific steps performed by the embodiment are described in detail in conjunction with table 1, table 2, and fig. 2 to 8.
The method comprises the steps of carrying out fault early warning on a certain wind turbine generator with a generator non-drive end bearing assembly damage fault in a certain wind power plant, selecting data collected by an SCADA system of the wind turbine generator in 2016-2017 for fault early warning when the generator non-drive end bearing assembly damage occurs in 2017.04.17, wherein the data sampling interval of the SCADA system is 5min, the data information is 16 months, the time range is 2016.01.0100: 00: 00-2017.04.3023: 55:00, the temperature measured at a generator non-drive end bearing temperature measuring point is selected as a target variable, and all parameters influencing the target variable value, such as other operating parameters of the generator, system parameters and the like, are taken as related variables. The dataset specific variables are shown in table 1:
TABLE 1 target variable and related variable of certain wind turbine of certain wind farm
The implementation data set of the wind turbine generator non-drive end bearing assembly fault early warning method in the embodiment is the 16-month operation data of the wind turbine generator, and the implementation steps of the method are as follows:
1) acquiring a running data set recorded in the fan SCADA system under a normal running state, wherein the data set comprises the temperature of a non-drive end bearing of a target variable generator and all related variables, and selecting data in the normal running state in the previous 12 months, namely 2016.01.0100: 00:00 to 2016.12.3123: 55:00 data to construct an initial training setThe last 4 months, i.e. 2017.01.0100: 00:00 to 2017.04.3023: 55:00Data construction initial test set
2) For the initial training setIn the method, all variables are subjected to data preprocessing, the preprocessing step comprises the elimination and interpolation of isolated outliers, the elimination and interpolation of data during the shutdown maintenance of the fan and the interpolation of missing values based on operation and maintenance records, 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 and the rated power is 1500kW, and a test set is subjected to data preprocessingAs a real-time running data set in online application, wherein training set and test set data before temperature pretreatment of a non-drive-end bearing of a target variable generator are shown in fig. 2, missing values and data during shutdown appear as 0 values, and the training set and test set data after pretreatment are shown in fig. 3;
3) selecting a DA-RNN model as a variable estimation model, selecting the length W of a sliding window to be 5, and the time in the sliding window to be twT, each time T is assigned by the design input attention mechanismwDependent variableFor target variableInfluence weight ofReconstructing correlated variablesThe encoder part is used as the input of an encoder, the encoder part is a plurality of LSTM units, the input of each LSTM unit is a reconstruction related variable of a moment in a sliding window, and the output of the encoder part is a hidden vector h; later designTime attention mechanism for distributing hidden vectors at different historical moments in sliding window to target variable at current momentInfluence weight ofObtaining a semantic vector c, integrating the semantic vector and a target variable historical value by using a linear regression model as the input of a decoder, wherein the decoder is partially composed of a plurality of LSTM units, and the output is the last moment t of the current momentT-1The estimated value of the target variable at the current moment is obtained by the linear functionTraining a model through a training set, and then inputting the test set to obtain a real-time estimation value of the test set, wherein a variable estimation result is shown in fig. 6, and a black circle is an actual operation 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 estimated value of the target variable, to obtain the estimated residual sequence of the training set, and calculating the mean value mu of the residual sequencetrainAnd standard deviation sigmatrain;
5) Performing multi-threshold setting based on the residual sequence of the training set, wherein the multi-threshold setting is the mean value mu of the estimated residual sequence of the training settrainPlus or minus k times standard deviation sigmatrainRespectively as the upper and lower limits of the threshold of the residual sequence, wherein the upper limit Ur(k)=μtrain+kσtrainLower limit of Lr(k)=μtrain-kσtrainWherein k is [1.5,2,2.5,3]]In fig. 7, the solid black line indicates a plurality of upper threshold limits, and the dashed black line indicates a plurality of lower threshold limits;
6) the test set is used as an on-line application phase data set, a data point d is operated in real time based on the test set, and an estimation residual value r of the actual measurement value of the data point d minus the model estimation value is obtained through a DA-RNN model trained in an off-line phased;
7) Selecting a value of k to determineDefining upper and lower limits of the threshold, if rdExceeds the corresponding threshold value of the current k value, namely is greater than the upper limit U of the threshold valuer(k) Or less than the lower threshold limit Lr(k) Calculating the number count1(k) of data points which are continuously over the threshold value before the data point d, and the percentage value count2 (k)% of the number of data points which are over the threshold value in the time range of one day before the data point d in the total number of data points in one day if rdAt the upper threshold Ur(k) And a lower threshold Lr(k) Meanwhile, two judgment results 0 are output;
8) setting multiple criteria, namely, combining the continuous overrun criterion with the percentage overrun criterion, setting a threshold parameter S (k) of the continuous overrun criterion and a threshold parameter P (k)%, where the parameters are set to be continuous overrun criterion and percentage overrun criterion, respectivelyWhen the condition count1(k) is more than or equal to S (k), the judgment result of the k-S (k) early warning strategy combination is marked as 1, when the condition count1(k) is more than or equal to S (k), the judgment result of the k-S (k) early warning strategy combination is marked as 0, when the condition count2(k) is more than or equal to P (k), the judgment result of the k-P (k) early warning strategy combination is marked as 1, and when the condition count2(k) is not more than P (k), the;
9) repeating the steps 7) and 8) for all other values of k), outputting two 0-1 judgment results for each value of k, and finally outputting a 0-1 sequence for the data point d, wherein the dimension of the sequence is 8 dimensions in the embodiment, when the number of 1 in the sequence is greater than 4, giving an alarm to the data point d, when the number of 1 in the sequence is less than or equal to 4, giving no alarm, and executing the operation on all data points in the test set, wherein fig. 8 is a final early warning result, a black star is a data point giving an alarm, a black circle is a data point at the time of fault confirmation, the earliest alarm time is 2017.03.2302: 00:00, and the judgment result of the time point is shown in the following table:
TABLE 2 earliest judgment result of 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 judgment result sequence is [0,1,1,1,1,1,0,1], the number of 1 is 6, so that the alarm is finally given at the moment, and the early warning is realized 24 days before the fault occurs.
The invention discloses a wind turbine generator key component fault early warning method which mainly comprises the steps 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 of a critical component fault early warning of the present invention, fig. 2, fig. 3 are data diagrams before and after preprocessing of 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 DA-RNN model selected in the present invention, fig. 5 is a schematic diagram of a decoder for adding a time attention mechanism to a DA-RNN model selected in the present invention, fig. 6 is a schematic diagram of a target variable estimation result of a variable estimation model in an embodiment of the present invention, fig. 7 is a schematic diagram of multi-threshold setting in an embodiment of the present invention, and fig. 8 is a diagram of an early warning result in an embodiment of the present invention.
The above-described embodiments are merely illustrative of the present invention, and although the best mode of the invention and the drawings are disclosed for illustrative purposes, those skilled in the art will appreciate that: various substitutions, changes and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the present invention should not be limited to the disclosure of the preferred embodiments and the accompanying drawings.
Claims (6)
1. A DA-RNN-based wind turbine generator key component fault early warning method 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 fan SCADA system under a normal operation state, selecting a temperature variable measured by a temperature measuring point of a component to be subjected to early warning in the SCADA system as a target variable y, and all variables related to the temperature of the component as related variables X, and constructing an initial training set
2) An off-line training phase based on an initial training setDesigning a data preprocessing flow, wherein the preprocessing step comprises the elimination and interpolation of isolated abnormal points, the elimination and interpolation of data during the shutdown maintenance of the fan based on operation and maintenance records, the interpolation of missing values, and the preprocessing of a training set [ X ]train,ytrain]Performing model training as input of a variable estimation model;
3) selecting a circular neural network DA-RNN model based on a double-attention machine system as a variable estimation model, selecting the length W of a sliding window, and expressing the time in the sliding window as twT, where T is the current time, and is entered by designMean force mechanism allocation at each time twDependent variableFor target variableInfluence weight ofReconstructing correlated variablesThe encoder part is used as the input of an encoder, the encoder part is a plurality of LSTM units, the input of each LSTM unit is a reconstruction related variable of a moment in a sliding window, and the output of the encoder part is a hidden vector h; designing time attention mechanism to distribute hidden vectors at different historical moments in sliding window to target variables at current momentInfluence weight ofObtaining a semantic vector c, integrating the semantic vector and a target variable historical value by using a linear regression model as the input of a decoder, wherein the decoder is partially composed of a plurality of LSTM units, and the output is the last moment t of the current momentT-1The estimated value of the target variable at the current moment is obtained by the 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, to obtain the estimated residual sequence of the training set, and calculating the mean value mu of the residual sequencetrainAnd standard deviation sigmatrain;
5) Performing multi-threshold setting based on training set residual sequence, and estimating the multi-threshold setting as the training setMean value μ of residual sequencetrainPlus or minus k times standard deviation sigmatrainRespectively as the upper and lower limits of the threshold of the residual sequence, wherein the upper limit Ur(k)=μtrain+kσtrainLower limit of Lr(k)=μtrain-kσtrain;
6) In the on-line application stage, based on the real-time operation data point d and the DA-RNN model trained in the off-line stage, the estimation residual value r of the model estimation value subtracted from the actual measurement value of the data point d is obtainedd;
7) Selecting a value of k, determining the upper and lower limits of the threshold value, if rdExceeds the corresponding threshold value of the current k value, namely is greater than the upper limit U of the threshold valuer(k) Or less than the lower threshold limit Lr(k) Calculating the number count1(k) of data points which are continuously over the threshold value before the data point d, and the percentage value count2 (k)% of the number of data points which are over the threshold value in the time range of one day before the data point d in the total number of data points in one day if rdAt the upper threshold Ur(k) And a lower threshold Lr(k) Meanwhile, two judgment results 0 are output;
8) setting multiple judgment criteria, wherein the multiple judgment criteria are continuous overrun judgment criteria combined with percentage overrun judgment criteria, setting a threshold parameter S (k) of the continuous overrun judgment criteria and a threshold parameter P (k)%) of the percentage overrun judgment criteria under a current k value, recording a judgment result of a k-S (k) early warning strategy combination as 1 when a condition count1(k) is greater than or equal to S (k), recording a judgment result as 0 when the condition is not met, recording a judgment result of a k-P (k) early warning strategy combination as 1 when a condition count2(k) is greater than or equal to P (k), and recording a judgment result as 0 when the condition is not met;
9) and repeating the step 7) and the step 8) for all other values of k to obtain a discrimination result sequence, and judging whether to give final alarm to the real-time data point d or not based on the discrimination result sequence.
2. The DA-RNN-based wind turbine generator key component fault early warning method as claimed in claim 1, wherein in the step 2), the preprocessing process of the off-line training comprises the following steps:
a) the isolated abnormal point is usually a recording error caused by the abnormality of the sensor, and is judged through an operation mechanism, wherein the judgment condition is as follows: for temperature variations, the value is greater than 150 degrees or less than 0 degrees; for the wind speed variable, the value is larger than the cut-in wind speed of the unit or smaller than the cut-in wind speed; for the power variable, the value is larger than the rated power of the unit or is a negative value; when the above conditions are satisfied, the data is judged to be isolated abnormal points and is removed.
b) Checking by combining with operation and maintenance records, when the wind turbine generator is in an operation, maintenance and repair period, the wind turbine generator is in a shutdown state, data recorded by the SCADA system is usually a 0 value or a system default value, and the value cannot represent a normal operation state of the wind turbine generator, so that data in the shutdown, maintenance and repair period are removed.
c) In order to ensure the time continuity of the data, the removed data and the missing value in 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 DA-RNN-based wind turbine generator key component fault early warning method as claimed in claim 1, wherein in the step 3), the model input is a related variable in a sliding windowWherein n is the number of the relevant variables, the input attention mechanism part inputs the relevant variables X, the hidden state output h and the memory unit output s of the encoder, and outputs the influence weight of the kth relevant variable on the target variableThe calculation process is as follows:
wherein v isen,Wen,UenFor the parameter to be learned, at time twBased on impact weightReconstructing the related variable to obtain the time twIs represented as:
the reconstructed vector is input as part of an encoder, which is composed of a plurality of LSTM units and outputs a hidden state h and a memory unit s at a time twUpdate of LSTM by forgetting gate ftInput door itOutput gate otThe decision, the update rule is as follows:
wherein Wf,Wi,Wo,Ws,bf,bi,bo,bsFor the parameter to be learned, σ is a sigmoid function, which represents multiplication of corresponding elements;
with time attention mechanismThe input is hidden state h, decoder hidden state output h 'and memory unit output s', and the weight of ith hidden state is outputThe calculation process is as follows:
wherein v isde,Wde,UdeReconstructing the hidden state for the parameter needing to be learned based on the weight beta to obtain a semantic vector, wherein the semantic vector is expressed as:
then, the linear regression model is used for integrating the semantic vector and the target variable historical value to obtain the input of the decoderExpressed as:
the decoder is also composed of LSTM units, forgetting gate ft', input door it', output gate otThe update rule of' is as follows:
wherein W'f,W′i,W′o,W′s,b′f,b′i,b′o,b′sParameters required to be learned;
finally, obtaining the estimated value of the target variable T at the current moment through a linear functionAs follows:
wherein WT,bTIs the parameter required to be learned.
4. The DA-RNN-based wind turbine generator critical component fault early warning method as claimed in claim 1, wherein in the step 5), it can be assumed that the training set residual sequence follows normal distribution, and based on normal distribution characteristics, the value of k in multi-threshold setting is [1.5,2,2.5,3 ].
5. The DA-RNN-based wind turbine generator key component fault early warning method as claimed in claim 1, wherein in the step 8), multiple criteria are determinedCorresponding to different abnormal characteristics, the continuous out-of-limit judgment criterion corresponds to the abnormal characteristics with continuous abnormal values, the percentage out-of-limit judgment criterion corresponds to the abnormal conditions with large data fluctuation, and the threshold parameter S of the continuous out-of-limit judgment criterionkAnd a percentage overrun criterion threshold parameter Pk% set toWherein C is a constant.
6. The DA-RNN-based wind turbine generator key component fault early warning method as claimed in claim 1, wherein in step 9), for the real-time data points d, two 0-1 judgment results are output for each k value, a 0-1 sequence is finally output, the dimensionality is 2k, when the number of 1 s in the sequence is greater than k, an alarm is given to the real-time data points, and when the number of 1 s 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 true CN111814849A (en) | 2020-10-23 |
CN111814849B 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) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112581031A (en) * | 2020-12-30 | 2021-03-30 | 杭州朗阳科技有限公司 | Method for realizing real-time monitoring of motor abnormity by Recurrent Neural Network (RNN) through C language |
CN113516273A (en) * | 2021-04-02 | 2021-10-19 | 中国船舶重工集团公司军品技术研究中心 | Fault prediction method for diesel engine supercharger for power generation |
CN114372504A (en) * | 2021-12-06 | 2022-04-19 | 燕山大学 | Wind turbine generator fault early warning method based on graph neural network |
CN114429249A (en) * | 2022-04-06 | 2022-05-03 | 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 |
CN114460481A (en) * | 2022-01-27 | 2022-05-10 | 重庆邮电大学 | Energy storage battery thermal runaway early warning method based on Bi-LSTM and attention mechanism |
Citations (11)
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 |
US20200150622A1 (en) * | 2018-11-13 | 2020-05-14 | Guangdong University Of Technology | Method for detecting abnormity in unsupervised industrial system based on deep transfer learning |
-
2020
- 2020-06-22 CN CN202010573207.3A patent/CN111814849B/en active Active
Patent Citations (11)
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 |
US20200150622A1 (en) * | 2018-11-13 | 2020-05-14 | Guangdong University Of Technology | Method for detecting abnormity in unsupervised industrial system based on deep transfer learning |
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 |
---|
杨婷婷;张蓓;吕游;邸小慧: "基于MSET的电站风机故障预警技术研究", 热能动力工程, vol. 32, no. 9 * |
胡瑾秋;张来斌;伊岩;蔡爽: "非正常工况下化工过程设备故障实时关联预警研究", 中国安全科学学报, vol. 26, no. 9 * |
黄小光;潘东浩;史晓鸣;王杏;王欣;: "风电机组齿轮箱系统的故障预警研究", 机电工程, no. 06 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112581031A (en) * | 2020-12-30 | 2021-03-30 | 杭州朗阳科技有限公司 | Method for realizing real-time monitoring of motor abnormity by Recurrent Neural Network (RNN) through C language |
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 |
CN113516273A (en) * | 2021-04-02 | 2021-10-19 | 中国船舶重工集团公司军品技术研究中心 | Fault prediction method for diesel engine supercharger for power generation |
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 |
CN114372504A (en) * | 2021-12-06 | 2022-04-19 | 燕山大学 | Wind turbine generator fault early warning method based on graph neural network |
CN114460481A (en) * | 2022-01-27 | 2022-05-10 | 重庆邮电大学 | Energy storage battery thermal runaway early warning method based on Bi-LSTM and attention mechanism |
CN114429249A (en) * | 2022-04-06 | 2022-05-03 | 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 |
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 |
Also Published As
Publication number | Publication date |
---|---|
CN111814849B (en) | 2024-02-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111814849A (en) | DA-RNN-based wind turbine generator key component fault early warning method | |
Udo et al. | Data-driven predictive maintenance of wind turbine based on SCADA data | |
CN111537219B (en) | Fan gearbox performance detection and health assessment method based on temperature parameters | |
CN112784373B (en) | Fault early warning method for wind turbine generator gearbox | |
CN111415070A (en) | Wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data | |
CN109583075B (en) | Permanent magnet direct-drive wind turbine service quality evaluation method based on temperature parameter prediction | |
Hwang et al. | SVM-RBM based predictive maintenance scheme for IoT-enabled smart factory | |
CN115578084A (en) | Wind turbine generator set frequency converter fault early warning method based on deep convolution self-encoder | |
CN111814848B (en) | Self-adaptive early warning strategy design method for temperature faults of wind turbine generator | |
CN109492866A (en) | A kind of distribution Running State intelligent evaluation method | |
CN107728059A (en) | A kind of pitch-controlled system state evaluating method | |
CN115222141A (en) | Multivariable time sequence abnormity detection method, system, medium, equipment and terminal | |
Kim et al. | Design of wind turbine fault detection system based on performance curve | |
CN115828466A (en) | Fan main shaft component fault prediction method based on wide kernel convolution | |
Shi et al. | Study of wind turbine fault diagnosis and early warning based on SCADA data | |
CN111396266A (en) | GBRT-based wind turbine generator bearing fault early warning method | |
Sarwar et al. | Time series method for machine performance prediction using condition monitoring data | |
Tutiv'en et al. | Wind turbine main bearing condition monitoring via convolutional autoencoder neural networks | |
CN116793666A (en) | Wind turbine generator system gearbox fault diagnosis method based on LSTM-MLP-LSGAN model | |
CN116226679A (en) | Wind turbine generator gearbox abnormality detection method considering similarity of running states of multiple units | |
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 | |
Souza et al. | Evaluation of data based normal behavior models for fault detection in wind turbines | |
Encalada-Dávila et al. | Wind turbine multi-fault detection based on SCADA data via an AutoEncoder | |
Rashid et al. | Anomaly Detection of Wind Turbine Gearbox based on SCADA Temperature Data using 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 | ||
GR01 | Patent grant |