CN114372504A - Wind turbine generator fault early warning method based on graph neural network - Google Patents
Wind turbine generator fault early warning method based on graph neural network Download PDFInfo
- Publication number
- CN114372504A CN114372504A CN202111478422.6A CN202111478422A CN114372504A CN 114372504 A CN114372504 A CN 114372504A CN 202111478422 A CN202111478422 A CN 202111478422A CN 114372504 A CN114372504 A CN 114372504A
- Authority
- CN
- China
- Prior art keywords
- data
- time
- temperature
- variables
- early warning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- 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
- G01—MEASURING; TESTING
- G01K—MEASURING TEMPERATURE; MEASURING QUANTITY OF HEAT; THERMALLY-SENSITIVE ELEMENTS NOT OTHERWISE PROVIDED FOR
- G01K13/00—Thermometers specially adapted for specific purposes
- G01K13/04—Thermometers specially adapted for specific purposes for measuring temperature of moving solid bodies
- G01K13/08—Thermometers specially adapted for specific purposes for measuring temperature of moving solid bodies in rotary movement
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
-
- 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
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Wind Motors (AREA)
Abstract
The invention belongs to the field of wind turbine state monitoring and fault early warning, and relates to a wind turbine fault early warning method based on a graph neural network, wherein S1, multivariate time sequence acquisition and data preprocessing are carried out; s2, decoupling the influence of the working condition change on the temperature variable to obtain decoupled temperature sensor data; s3, inputting the health data after the decoupling processing into a space-time graph network, and advancing space-time association characteristics; s4, setting a threshold value according to the verification set; s5, inputting the online data into the model, calculating an abnormal score, and judging whether to perform fault early warning according to a threshold value; according to the invention, the influence of the working condition change on the temperature state parameters is decoupled, and the dynamic time-space characteristics among different temperature sensor parameters are effectively extracted by using the graph neural network, so that the fault early warning accuracy and reliability are improved.
Description
Technical Field
The invention belongs to the field of wind turbine state monitoring and fault early warning, and relates to a wind turbine fault early warning method based on a graph neural network.
Background
In recent years, with the increase of the installation number of wind turbines, the wind turbines and key parts thereof are easy to break down due to complex operation conditions and severe working environments, and even the wind turbines stop in severe cases. This often results in huge economic losses and bad social impact. Therefore, the early warning of the unit fault is realized, and the early warning method has important significance. According to the fault development trend, the potential faults can be discovered as soon as possible, the optimal maintenance strategy can be formulated, the fault rate is reduced, the operation and maintenance cost is reduced, and major faults can be prevented through fault early warning, so that major property loss is avoided, and the personal and equipment safety is guaranteed.
The current wind turbine generator system fault monitoring mainly comprises the steps that a vibration sensor and a temperature sensor are arranged through a generator system transmission system, and the running state of a wind turbine generator is monitored through a monitoring means based on vibration mode analysis and a detection means based on infrared scanning. However, the failure root of the unit component is complex, the detection difficulty is high, and the fault of a certain component of the unit cannot be timely and accurately warned. In recent years, a Supervisory Control and Data Acquisition (SCADA) system has been widely deployed in a wind farm, the system includes hundreds of sensor parameters, and the parameters related to the present invention mainly include two categories: one type is temperature state parameters which are numerous and are related to the health state of each component (such as a main bearing, a gearbox, a generator and the like) of the unit; the other is an environmental condition parameter, i.e. a parameter that affects the temperature of the component, such as wind speed, power, ambient temperature, etc.
The subsystems of the unit are closely related and interdependent, and have a plurality of synergistic effects, and the health state of the unit and the components thereof is represented by a plurality of monitoring parameters. When the operation state of the unit is abnormal or fails, a plurality of state parameters may change simultaneously, and the correlation between the parameters may be affected. At present, most of methods for early warning faults of key parts of a wind turbine generator manually select a certain characteristic monitoring parameter according to historical experience to carry out research. However, the wind turbine generator is a strong coupling system with multiple subsystems working in a cooperative mode, the information content of a single monitoring parameter is low, and the abnormal state of the system is difficult to reflect comprehensively. In addition, the prior art can not effectively model and mine the dynamic time-space relationship among different sensor parameters of the SCADA, so that the fault early warning accuracy and reliability are low. Therefore, the time-space correlation among different sensor parameters is mined from the analysis view of the graph data, the deep learning fault early warning based on the time-space graph neural network is established, and the reliability of the fault early warning is improved.
Disclosure of Invention
The invention aims to solve the defects of the problems and provides a method for effectively predicting the faults of the wind turbine generator in advance and early warning the faults based on a graph neural network.
In order to solve the technical problems, the technical scheme adopted by the invention is as follows:
a wind turbine generator fault early warning method based on a graph neural network comprises the following steps:
s1, multivariate time series acquisition and data preprocessing: acquiring multivariate sensor time sequence historical data under the health state of a wind turbine generator from a wind power plant state monitoring and data acquisition SCADA system, cleaning the data to remove abnormal data, and screening required working condition variables and temperature state variables;
s2, decoupling the influence of the working condition change on the temperature variable to obtain decoupled temperature sensor data: taking the working condition variable obtained by screening in the S1 as input, taking the temperature state variable as output, establishing a multi-input multi-output regression model, further calculating the difference value of the original temperature state variable and the temperature state variable predicted value obtained by the regression model as state variable time series data after decoupling the working condition change, and dividing the state variable time series data into a training set and a verification set;
s3, inputting the health data after the decoupling processing into a space-time graph network, and leading the space-time correlation characteristics: constructing a prediction model based on a space-time graph convolutional network, firstly extracting space correlation characteristics among different sensor variables in decoupled state variable time sequence data by using an attention mechanism, and calculating to obtain an adjacency matrix representing the space correlation of the sensors; meanwhile, inputting the decoupled state variable time series data into a time convolution module, extracting the characteristics of the time dimension, inputting the characteristics of the time dimension and the adjacency matrix into a graph convolution module, and extracting the characteristics of the space dimension; training a space-time graph convolutional network prediction model by using a training set;
s4, setting a threshold value according to the verification set: inputting the verification set into a trained time-space diagram convolutional network prediction model, calculating predicted values and residual errors of true values of all variables in the verification set, selecting the maximum residual error as an abnormal score, and setting a threshold value of fault early warning according to statistical distribution of the abnormal score;
s5, inputting the online data into the model, calculating an abnormal score, and judging whether to perform fault early warning according to a threshold value: acquiring online real-time monitoring multivariable time data from a wind turbine generator, inputting the multivariable time data into a multi-input multi-output regression model, calculating to obtain decoupled temperature state data, inputting the decoupled temperature state data into a trained space-time diagram convolution prediction model, calculating an abnormal score according to a residual error between a predicted value and a true value, and comparing the abnormal score with a preset threshold value; and when the abnormal score is larger than the threshold value, sending out fault early warning to the fan component.
The technical scheme of the invention is further improved as follows: in S1, the method includes the steps of:
s11, performing outlier detection on the original SCADA data by using an outlier factor detection algorithm, and removing data which do not accord with physical significance;
s12, screening temperature state variables related to the wind turbine state and working condition variables influencing the temperature to perform next modeling, wherein the working condition variables comprise environment temperature, wind speed, power and rotor rotating speed, and the temperature state variables comprise main bearing temperature, gear box temperature, generator bearing temperature and the like.
The technical scheme of the invention is further improved as follows: in S2, the method includes the steps of:
s21, taking the working condition parameters as independent variables, taking all temperature state variables as dependent variables, establishing a multi-input and multi-output regression decoupling model by using a fully-connected neural network, and calculating to obtain regression model residual errors as the temperature state variables influenced by decoupling working conditions;
and S22, establishing time series data for the decoupled temperature variables by adopting a sliding window, and dividing the data into a training set and a verification set sample according to months, wherein the first 80% is used as the training set, and the remaining 20% is used as the verification set.
The technical scheme of the invention is further improved as follows: in S3, the method includes the steps of:
s31, training set data X is belonged to RN×TThe input is gated around a unit, where N is the number of variables and T is the time step, which layer sequentially computes the hidden state corresponding to each timestamp T. The last hidden state R is used as a representation of the entire time series and the weight matrix W is calculated by a self-attention mechanism. Calculating an adjacency matrix A of the sensor space coupling diagram according to the weight matrix W, wherein A is 0.5WT；
S32, training set data X is belonged to RN×TRespectively inputting sigmoid as one-dimensional causal convolution of an activation function and tanh as one-dimensional causal convolution of the activation function, and calculating to obtain a characteristic h on the time dimensiont(X) the calculation formula is as follows:
wherein U and V are convolution kernel parameters, b and c are bias parameters,
an operation representing multiplication of corresponding elements;
s33, h obtained in S32t(X) the adjacent matrix a obtained in S31 is input to the chebyshev map convolution layer, and the output is calculated as follows:
X′∈Rm×n×k，
wherein m is the number of adjacent matrixes, n is the number of variables, and k is the number of convolution kernels;
and S34, inputting the X' obtained in the step S33 into two one-dimensional convolutional layers, converting the input channel dimension into a required output dimension, setting the average absolute error as a loss function, and finishing model training.
Due to the adoption of the technical scheme, the invention has the technical progress that:
the invention provides a wind turbine generator fault early warning method based on a graph neural network. In addition, the method can adaptively learn the relevance of different temperature parameter time by designing the self-adaptive construction dynamic graph adjacency matrix based on the self-attention mechanism, and simultaneously fully extract the space characteristics by using the graph convolution network and extract the time characteristics by designing the time convolution module, thereby realizing the effective extraction of the space characteristics, establishing a global state monitoring model based on temperature variables, realizing the early reliable early warning of unit faults, being beneficial to timely processing and maintaining the unit and avoiding the deep damage of the unit and key parts thereof.
Drawings
FIG. 1 is a schematic flow diagram of the present invention.
FIG. 2 is a schematic diagram of the decoupling model of the present invention.
Fig. 3 is a schematic diagram of a space-time diagram network according to the present invention.
FIG. 4 is a graphical representation of the original 24-dimensional temperature variation in a comparison plot of the decoupling model results of the present invention.
FIG. 5 is a schematic diagram of the results of the decoupling model of the present invention versus the temperature variables in the graph after the decoupling analysis.
FIG. 6 is a graph of the test set anomaly score results of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the following examples:
FIG. 1 is a flow chart of a wind turbine generator fault early warning method based on a graph neural network, and the method comprises five parts, namely obtaining a multivariate time sequence, preprocessing data, decoupling the influence of working condition change on temperature variables, obtaining decoupled temperature sensor data, inputting the decoupled health data into an air-map network, extracting space-time correlation characteristics, setting a threshold value according to a verification set, inputting online data into a model, calculating an abnormal score, and judging whether fault early warning is performed or not according to a valve group.
The decoupling model is schematically shown in fig. 2, the operating condition parameters are used as independent variables, all temperature state variables are used as dependent variables, a multi-input and multi-output regression decoupling model is established by using a fully-connected neural network, and regression model residuals are obtained through calculation and used as temperature state variables influenced by decoupling operating conditions. The working condition parameters are 4-dimensional in total and comprise wind speed, ambient temperature, power and rotor rotating speed; the temperature parameters in the decoupling model are all temperature variables, and the total 24 dimensions are respectively as follows: generator bearing temperature 1, generator bearing temperature 2, generator stator winding temperature 1, generator stator winding temperature 2, generator stator winding temperature 3, hydraulic group oil temperature, gearbox bearing temperature on the high speed shaft, nacelle temperature, high speed transformer temperature 1, high speed transformer temperature 2, high speed transformer temperature 3, grid side inverter temperature, top nacelle controller temperature, hub controller temperature, VCP board temperature, separator ring room temperature, head cone temperature, VCP blocking coil temperature, IGBT-driver temperature on rotor side inverter 1, IGBT-driver temperature on rotor side inverter 2, IGBT-driver temperature on rotor side inverter 3, VCP cooling water temperature, bus section temperature.
Fig. 3 is a schematic diagram of a space-time diagram network. Making the training set data X be equal to RN×TThe input is gated around a unit, where N is the number of variables and T is the time step, which layer sequentially computes the hidden state corresponding to each timestamp T. The last hidden state R is used as a representation of the entire time series and the weight matrix W is calculated by a self-attention mechanism. Calculating an adjacency matrix A of the sensor space coupling diagram according to the weight matrix W, wherein A is 0.5WT。
Then training set data X is belonged to RN×TRespectively inputting sigmoid as a one-dimensional causal convolution of the activation function and tanh as a one-dimensional causal convolution of the activation function,calculating to obtain the characteristic h of the time dimensiont(X) the formula for calculation is:
where U and V are convolution kernel parameters, b and c are bias parameters,
representing the operation of multiplying corresponding elements.
H is to bet(X) inputting the Chebyshev chart convolution layer with the adjacency matrix A, and calculating to obtain the output X' belonging to Rm×n×kWherein m is the number of adjacent matrixes, n is the number of variables, and k is the number of convolution kernels.
Inputting X' into the two one-dimensional convolution layers, converting the input channel dimension into the required output dimension, setting the average absolute error as a loss function, and completing model training. The maximum value of the training set residuals is calculated and set as the threshold.
In the online fault early warning part, online real-time monitoring multivariable time data are obtained from a wind turbine generator, the multivariable time data are input into a multi-input multi-output regression model firstly, decoupled temperature state data are obtained through calculation, then the decoupled temperature state data are input into a trained spatio-temporal graph convolution prediction model, the maximum value of a residual error between a predicted value and a true value is calculated to be an abnormal score, and the abnormal score is compared with a preset threshold value; and when the abnormal score is larger than the threshold value, sending out fault early warning to the fan component.
Fig. 4 and 5 are graphs comparing results of decoupling models, fig. 4 is an original 24-dimensional temperature variable, and fig. 5 is a temperature variable after decoupling analysis, and it can be seen that the non-stationarity of the variable is eliminated to a great extent.
On-line fault early warning is shown in fig. 6, a yellow line is a threshold, a blue line is an abnormal score of a test set every day, and a fault occurs in the last day of the test set. It can be seen that the 2016 month 7 and day 10 anomaly score exceeds a threshold and continues to rise, thus allowing for effective early warning of gearbox failure 8 days in advance.
The above-mentioned embodiments are merely illustrative of the preferred embodiments of the present invention, and do not limit the scope of the present invention, and various modifications and improvements of the technical solution of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the appended claims.
Claims (4)
1. A wind turbine generator fault early warning method based on a graph neural network is characterized by comprising the following steps:
s1, multivariate time series acquisition and data preprocessing: acquiring multivariate sensor time sequence historical data under the health state of a wind turbine generator from a wind power plant state monitoring and data acquisition SCADA system, cleaning the data to remove abnormal data, and screening required working condition variables and temperature state variables;
s2, decoupling the influence of the working condition change on the temperature variable to obtain decoupled temperature sensor data: taking the working condition variables obtained by screening in the step S1 as input, taking the temperature state variables as output, establishing a multi-input multi-output regression model, further calculating the difference value of the original temperature state variables and the predicted values of the temperature state variables obtained by the regression model as state variable time series data after decoupling the working condition changes, and dividing the state variable time series data into a training set and a verification set;
s3, inputting the health data after the decoupling processing into a space-time graph network, and leading the space-time correlation characteristics: constructing a prediction model based on a space-time graph convolutional network, firstly extracting space correlation characteristics among different sensor variables in decoupled state variable time sequence data by using an attention mechanism, and calculating to obtain an adjacency matrix representing the space correlation of the sensors; meanwhile, inputting the decoupled state variable time series data into a time convolution module, extracting the characteristics of the time dimension, inputting the characteristics of the time dimension and the adjacency matrix into a graph convolution module, and extracting the characteristics of the space dimension; training a space-time graph convolutional network prediction model by using a training set;
s4, setting a threshold value according to the verification set: inputting the verification set into a trained time-space diagram convolutional network prediction model, calculating predicted values and residual errors of true values of all variables in the verification set, selecting the maximum residual error as an abnormal score, and setting a threshold value of fault early warning according to statistical distribution of the abnormal score;
s5, inputting the online data into the model, calculating an abnormal score, and judging whether to perform fault early warning according to a threshold value: acquiring online real-time monitoring multivariable time data from a wind turbine generator, inputting the multivariable time data into a multi-input multi-output regression model, calculating to obtain decoupled temperature state data, inputting the decoupled temperature state data into a trained space-time diagram convolution prediction model, calculating an abnormal score according to a residual error between a predicted value and a true value, and comparing the abnormal score with a preset threshold value; and when the abnormal score is larger than the threshold value, sending out fault early warning to the fan component.
2. The wind turbine generator fault early warning method based on the graph neural network as claimed in claim 1, wherein: in S1, the method includes the steps of:
s11, performing outlier detection on the original SCADA data by using an outlier factor detection algorithm, and removing data which do not accord with physical significance;
s12, screening temperature state variables related to the wind turbine state and working condition variables influencing the temperature to perform next modeling, wherein the working condition variables comprise environment temperature, wind speed, power and rotor rotating speed, and the temperature state variables comprise main bearing temperature, gear box temperature, generator bearing temperature and the like.
3. The wind turbine generator fault early warning method based on the graph neural network as claimed in claim 1, wherein: in S2, the method includes the steps of:
s21, taking the working condition parameters as independent variables, taking all temperature state variables as dependent variables, establishing a multi-input and multi-output regression decoupling model by using a fully-connected neural network, and calculating to obtain regression model residual errors as the temperature state variables influenced by decoupling working conditions;
and S22, establishing time series data for the decoupled temperature variables by adopting a sliding window, and dividing the data into a training set and a verification set sample according to months, wherein the first 80% is used as the training set, and the remaining 20% is used as the verification set.
4. The wind turbine generator fault early warning method based on the graph neural network as claimed in claim 1, wherein: in S3, the method includes the steps of:
s31, training set data X is belonged to RN×TInputting a gating cycle unit, wherein N is a variable number, T is a time step, and the layer calculates hidden states corresponding to each timestamp T in sequence; using the last hidden state R as the representation of the whole time sequence, calculating a weight matrix W through a self-attention mechanism, and calculating an adjacent matrix A of the sensor space coupling diagram according to the weight matrix W, wherein A is 0.5WT；
S32, training set data X is belonged to RN×TRespectively inputting sigmoid as one-dimensional causal convolution of an activation function and tanh as one-dimensional causal convolution of the activation function, and calculating to obtain a characteristic h on the time dimensiont(X) the formula for calculation is:
wherein U and V are convolution kernel parameters, b and c are bias parameters,
an operation representing multiplication of corresponding elements;
s33, h obtained in S32t(X) the output obtained by inputting the chebyshev map convolution layer and the adjacency matrix a obtained in S31 is calculated as:
X′∈Rm×n×k，
wherein m is the number of adjacent matrixes, n is the number of variables, and k is the number of convolution kernels;
and S34, inputting the X' obtained in the step S33 into two one-dimensional convolutional layers, converting the input channel dimension into a required output dimension, setting the average absolute error as a loss function, and finishing model training.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111478422.6A CN114372504A (en) | 2021-12-06 | 2021-12-06 | Wind turbine generator fault early warning method based on graph neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111478422.6A CN114372504A (en) | 2021-12-06 | 2021-12-06 | Wind turbine generator fault early warning method based on graph neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114372504A true CN114372504A (en) | 2022-04-19 |
Family
ID=81140528
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111478422.6A Pending CN114372504A (en) | 2021-12-06 | 2021-12-06 | Wind turbine generator fault early warning method based on graph neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114372504A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114692341A (en) * | 2022-04-22 | 2022-07-01 | 桂林电子科技大学 | Fan fault prediction method based on fusion space-time diagram ConvLSTM model and storage medium |
| CN114818515A (en) * | 2022-06-24 | 2022-07-29 | 中国海洋大学 | Multidimensional time sequence prediction method based on self-attention mechanism and graph convolution network |
| CN115453356A (en) * | 2022-09-06 | 2022-12-09 | 上海电力大学 | Power equipment running state monitoring and analyzing method, system, terminal and medium |
| CN115905980A (en) * | 2022-12-09 | 2023-04-04 | 马鞍山玖帆科技有限公司 | Equipment monitoring system for wind power energy and monitoring method thereof |
| CN116382223A (en) * | 2023-06-02 | 2023-07-04 | 山东鲁能控制工程有限公司 | Thermal power generating unit monitoring system based on DCS |
| CN116484751A (en) * | 2023-06-21 | 2023-07-25 | 北京尚文汇通能源科技有限公司 | Fatigue life assessment method and device for wind turbine generator components |
| WO2023221701A1 (en) * | 2022-05-16 | 2023-11-23 | 北京火山引擎科技有限公司 | Multivariable time sequence processing method and apparatus, device and medium |
-
2021
- 2021-12-06 CN CN202111478422.6A patent/CN114372504A/en active Pending
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114692341A (en) * | 2022-04-22 | 2022-07-01 | 桂林电子科技大学 | Fan fault prediction method based on fusion space-time diagram ConvLSTM model and storage medium |
| CN114692341B (en) * | 2022-04-22 | 2024-03-19 | 桂林电子科技大学 | Fan fault prediction method based on fused time-space diagram ConvLSTM model and storage medium |
| WO2023221701A1 (en) * | 2022-05-16 | 2023-11-23 | 北京火山引擎科技有限公司 | Multivariable time sequence processing method and apparatus, device and medium |
| CN114818515A (en) * | 2022-06-24 | 2022-07-29 | 中国海洋大学 | Multidimensional time sequence prediction method based on self-attention mechanism and graph convolution network |
| CN115453356A (en) * | 2022-09-06 | 2022-12-09 | 上海电力大学 | Power equipment running state monitoring and analyzing method, system, terminal and medium |
| CN115453356B (en) * | 2022-09-06 | 2023-05-30 | 上海电力大学 | Power equipment operation state monitoring and analyzing method, system, terminal and medium |
| CN115905980A (en) * | 2022-12-09 | 2023-04-04 | 马鞍山玖帆科技有限公司 | Equipment monitoring system for wind power energy and monitoring method thereof |
| CN116382223A (en) * | 2023-06-02 | 2023-07-04 | 山东鲁能控制工程有限公司 | Thermal power generating unit monitoring system based on DCS |
| CN116382223B (en) * | 2023-06-02 | 2023-08-01 | 山东鲁能控制工程有限公司 | Thermal power generating unit monitoring system based on DCS |
| CN116484751A (en) * | 2023-06-21 | 2023-07-25 | 北京尚文汇通能源科技有限公司 | Fatigue life assessment method and device for wind turbine generator components |
| CN116484751B (en) * | 2023-06-21 | 2023-09-05 | 北京尚文汇通能源科技有限公司 | Fatigue life assessment method and device for wind turbine generator components |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114372504A (en) | Wind turbine generator fault early warning method based on graph neural network | |
| CN106682814B (en) | Wind turbine generator fault intelligent diagnosis method based on fault knowledge base | |
| CN111537219B (en) | Fan gearbox performance detection and health assessment method based on temperature parameters | |
| Kim et al. | Use of SCADA data for failure detection in wind turbines | |
| CN110378042B (en) | Wind turbine generator gearbox oil temperature anomaly detection method and system based on SCADA data | |
| CN111380686A (en) | XGboost algorithm model-based fan main bearing fault monitoring and diagnosis method | |
| Hu et al. | Using domain knowledge features for wind turbine diagnostics | |
| CN111720296B (en) | Steam feed pump equipment abnormal state early warning method based on self-encoder | |
| CN112598172A (en) | Wind turbine bearing temperature early warning method | |
| CN111103137A (en) | Wind turbine gearbox fault diagnosis method based on deep neural network | |
| CN107291991A (en) | A kind of Wind turbines early defect method for early warning based on dynamic network mark | |
| CN115329986A (en) | Wind turbine generator anomaly detection and positioning method based on interpretable graph neural network | |
| Abbasi et al. | Development of predictive maintenance interface using multiple linear regression | |
| CN114841580A (en) | Generator fault detection method based on hybrid attention mechanism | |
| CN116609055A (en) | Method for diagnosing wind power gear box fault by using graph convolution neural network | |
| Wang et al. | Wind turbine fault detection and identification through self-attention-based mechanism embedded with a multivariable query pattern | |
| CN113313365A (en) | Degradation early warning method and device for primary air fan | |
| CN117232809A (en) | Fan main shaft fault pre-diagnosis method based on DEMATEL-ANP-CRITIC combined weighting | |
| Gao et al. | An Overview on fault diagnosis, prognosis and resilient control for wind turbine systems. Processes. 2021; 9: 300 | |
| Tutiv'en et al. | Wind turbine main bearing condition monitoring via convolutional autoencoder neural networks | |
| CN115687864A (en) | XGboost-KDE-based fan main bearing fault early warning method | |
| CN115578084A (en) | Wind turbine generator set frequency converter fault early warning method based on deep convolution self-encoder | |
| CN114841061A (en) | Wind power gear box health assessment method and system integrating timing sequence information | |
| Zhang | Comparison of data-driven and model-based methodologies of wind turbine fault detection with SCADA data | |
| Song et al. | Anomaly detection of wind turbine generator based on temporal information |
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 |