CN112834211A - Fault early warning method for transmission system of wind turbine generator - Google Patents
Fault early warning method for transmission system of wind turbine generator Download PDFInfo
- Publication number
- CN112834211A CN112834211A CN202011641131.XA CN202011641131A CN112834211A CN 112834211 A CN112834211 A CN 112834211A CN 202011641131 A CN202011641131 A CN 202011641131A CN 112834211 A CN112834211 A CN 112834211A
- Authority
- CN
- China
- Prior art keywords
- data
- wind turbine
- turbine generator
- model
- 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
- 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
- G01M13/02—Gearings; Transmission mechanisms
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Wind Motors (AREA)
Abstract
A wind turbine generator transmission system fault early warning method comprises the steps of firstly obtaining wind turbine generator SCADA operation data, selecting continuously-operated health data, and finishing relevant preprocessing on the data; then, constructing a long-short term memory self-coding model, and training the model by using a training set and a test set; and finally, verifying the reconstruction capability of the model by using a verification set, calculating the residual error between the reconstruction data and the original data, setting a threshold control line by adopting a kernel density estimation method, performing fault early warning when the reconstruction error is higher than a threshold, and judging the specific part with the fault according to the size of the reconstruction error and the time when the reconstruction error reaches the threshold. According to the invention, the gate control unit of the long-term and short-term memory network is combined with the denoising self-coding network, so that the internal space-time correlation of data can be effectively captured, the characteristic information of data and the characteristic information of data in a time dimension can be better mined, and the reliability of fault early warning can be effectively improved.
Description
Technical Field
The invention relates to a wind turbine generator transmission system fault early warning method.
Background
With the rapid development of social civilization, the demand of human beings for energy is gradually increased, and the solution of environmental problems caused by fossil energy consumption is urgent, and the active exploration and the full utilization of renewable energy sources become important problems facing countries in the world. Wind power, a clean, renewable energy source, has been developed and utilized worldwide in recent years on a large scale. According to data statistics, the accumulated installed total amount and the annual installed amount of China respectively account for 35% and 37% of the world, and the accumulated installed total amount and the annual installed amount occupy the most important positions in the whole wind power industry.
Wind turbine generators are generally distributed in coastal areas, mountainous areas and other areas with abundant wind energy but bad natural conditions, and various faults are easy to occur when the wind turbine generators continuously run for a long time. The transmission system is used as an important transmission device of the wind turbine generator, and works under complex working conditions of low speed, heavy load, alternating load effect, strong gust impact and the like for a long time, so that the fault occurrence rate is high, and the economic benefit and the social benefit of wind power are seriously influenced. Therefore, the healthy operation of the transmission system of the wind turbine generator is ensured, the early warning of the impending faults can be carried out, the economic loss of the wind power industry can be greatly reduced, and the utilization rate of wind energy is improved.
A Supervisory Control and Data Acquisition (SCADA) system is a monitoring system integrated in a wind turbine generator system without secondary installation, and hundreds of digital sensors monitor the operation condition of the whole wind turbine generator system in real time, and generate massive monitoring Data along with the operation of the wind turbine generator system. The method is an effective way for solving the health monitoring problem of the transmission system of the wind turbine generator by using massive data acquisition and monitoring control system operation data to establish a network model based on data driving. The operation data of the data acquisition and monitoring control system is actually time sequence data recorded by a plurality of sensors, and has the characteristics of space-time correlation, nonlinearity, strong coupling and a large amount of noise.
Therefore, how to develop a new method capable of accurately early warning the fault of the wind turbine generator according to the characteristics of data acquisition and monitoring of the control system, and reduce the operation and maintenance cost of the wind power industry is a problem to be solved urgently by researchers in the field.
Disclosure of Invention
The invention provides a wind turbine generator transmission system fault early warning method based on a long-short term memory self-coding network, and aims to solve the problems of complex coupling relations of all parts of a current wind turbine generator transmission system and time dependency of each dimension of data in operation data of a data acquisition and monitoring control system. According to the invention, the deep learning network model is established by using the data of the data acquisition and monitoring control system, on one hand, the wind turbine generator operation data obtained by monitoring the existing data acquisition and monitoring control system of the wind turbine generator is used, so that the economic cost of secondarily installing the sensor is saved; on the other hand, the deep neural network is established according to the multivariable high-dimensional characteristic of the operation data of the data acquisition and monitoring control system, and the early warning can be better carried out on the faults to be generated.
In order to achieve the purpose, the invention establishes a long-term and short-term memory self-coding network model based on data of a wind turbine generator data acquisition and monitoring control system, performs data reconstruction by selecting proper characteristics, obtains a fault early warning threshold value of each part according to a reconstruction error, and achieves fault early warning when the reconstruction error exceeds the threshold value.
The method comprises the following steps:
s1: acquiring running data of a wind turbine generator data acquisition and monitoring control system, and selecting health state data of continuous running of the wind turbine generator;
s2: in order to reduce the redundancy and the feature dimension of the data of the wind turbine generator data acquisition and monitoring control system, a correlation coefficient method is used for carrying out feature screening on the screened health state data of the wind turbine generator;
s3: smoothing the selected data by adopting an exponential weighted moving average method, and taking the processed data as a training set and a verification set of the model;
s4: constructing a long-short term memory self-coding network model, and training the model by using a training set to complete data coding and reconstruction learning;
s5: inputting the verification set into a trained long-short term memory self-coding network model, calculating a reconstruction residual error between reconstruction data and original input data, and setting a threshold control line by adopting a kernel density estimation method;
s6: and inputting data of the real-time operation of the wind turbine generator into the trained long-short term memory self-coding network model to calculate a reconstruction error, performing fault early warning when the reconstruction error is higher than a threshold value, and judging the specific component with the fault according to the size of the reconstruction error and the time when the reconstruction error reaches the threshold value.
Further, in step S1, the health status data is data of a data acquisition and monitoring control system acquired by the unit in a normal operation state, and the specific step of selecting the continuous health status data includes:
s1.1: collecting data of a data acquisition and monitoring control system of the wind turbine generator set which continuously operates for more than half a year, and screening out data of the wind turbine generator set in a normal operation state according to state codes in the data of the data acquisition and monitoring control system;
s1.2: and according to the time information of the wind turbine generator operation data, checking the continuity of the wind turbine generator operation data, and selecting healthy and continuous wind turbine generator operation data as a training set and a verification set.
Further, in step S2, the specific steps of the feature screening process include:
s2.1: determining characteristic data related to the running state of each component of a wind turbine generator transmission system, wherein the characteristic data comprises main bearing temperature, gearbox oil temperature, gearbox bearing temperature, gearbox oil pump pressure, hydraulic system pressure, generator rotating speed, generator bearing temperature and the like;
s2.2: by correlation coefficient methodCalculating a linear correlation coefficient between the variables, calculating a correlation coefficient r for two variables X and Y asWherein N is the number of samples, and X and Y are different variables; and then selecting the first variables with larger values of r as the input of the long-short term memory self-coding model according to the size of the correlation coefficient r.
Further, in step S4, the model mainly includes an input layer and a plurality of hidden layers and an output layer, the input layer and the output layer have the same dimension, wherein each hidden layer is composed of long-short term memory neural units, and the model is constructed by the following steps:
s4.1: determining a training set and a verification set, and determining the dimensions of an input layer and an output layer according to the screened data characteristics;
s4.2: training a model by using a single-layer neural network, and determining the optimal number of the first hidden layer neural units by a method of gradually increasing the neural units;
s4.3: under the premise of determining the number of the first hidden layer neural units, gradually increasing the depth of the network model in a mode of gradually decreasing the number of the neural units layer by layer;
s4.4: training the model by using an Adam algorithm, and determining parameters of the constructed long-short term memory self-coding network model;
s4.5: a denoising masking layer is added behind an input layer of the network model to enhance the anti-noise capability of the model and enable the model to have stronger robustness;
s4.6: and verifying the model by using the verification set, and saving the model when the model is determined to be capable of well reconstructing input data.
Further, in step S5, the calculating a reconstruction error between the original data and the reconstructed data and calculating a threshold control line includes:
s5.1: inputting the verification set Y into the trained long-short term memory self-coding network model, and calculating to obtain output reconstruction dataAnd calculating a reconstruction error E, the formula is as follows:
s5.2: the Robust Mahalanobis Distance (RMD) is adopted to complete the calculation of the monitoring index sequence h, and the kth sample EkIts corresponding monitoring index value hkThe calculation formula is as follows:
in the formula (I), the compound is shown in the specification,is a robust measure of central tendency, i.e., median; MCD-1Is an inverse covariance matrix calculated from a sample by a minimum covariance determination estimation method, and T represents transposition operation.
S5.3: determining a probability distribution function p (h) of a monitoring index sequence h by using a Kernel Density Estimation (KDE), wherein the calculation formula is as follows:
in the formula, σ is a kernel function broadband coefficient, K (·) is a kernel function, and N is the length of the monitoring index sequence h. Determining the value of sigma according to multiple experiments, and using Gaussian kernel as kernel functionWhere g is a function variable.
S5.4: and finally, calculating a corresponding monitoring threshold value d according to a probability density distribution function p (h) of the monitoring index sequence h obtained by estimation:
in the formula, α represents a confidence level.
Compared with the prior art, the self-coding network model is constructed based on the long-short term memory neural network aiming at the characteristics of nonlinearity, space-time relevance and large noise of the data of the wind turbine generator data acquisition and monitoring control system, so that on one hand, the space-time relevance of the data can be solved, and the characteristic information of the data among the data and the characteristic information of the data in the time dimension can be better mined; on the other hand, by introducing the denoising masking layer (Dropout) behind the input layer, the noise resistance of the model can be well improved, the robustness of the model is enhanced, and the input data can be better reconstructed.
The invention uses the data of the wind turbine generator data acquisition and monitoring control system, does not need to install a sensor for the second time, and saves the operation and maintenance cost. And the training model adopts an unsupervised learning strategy, only needs data in the health state of the unit, does not need fault marking data to participate in model training, and greatly reduces the marking cost of the data and the training cost of the model.
The method is used for establishing the model, the early warning can be carried out on the faults to be generated for hours in advance through engineering data verification, the specific parts to be generated with the faults are judged, guidance is provided for operation and maintenance personnel to maintain the unit, and economic loss caused by long-time shutdown can be greatly reduced.
Drawings
FIG. 1 is an off-line training flow chart of a wind turbine generator transmission system fault early warning method;
FIG. 2 is a flow chart of the detailed construction steps of a long-short term memory self-coding network model;
FIG. 3 is a block diagram of a long-short term memory self-coding network model;
FIG. 4 is a flow chart of a wind turbine generator transmission system fault early warning method of the present invention;
FIG. 5 is a verification graph based on engineering data using the method of the present invention.
Detailed Description
The core of the wind turbine transmission system fault early warning method is to establish a long-term and short-term memory self-coding network model based on wind turbine data acquisition and monitoring control system data, reconstruct the data by selecting proper characteristics, obtain a fault early warning threshold of each component according to reconstruction errors, and realize fault early warning when the reconstruction errors exceed the thresholds.
As shown in fig. 1, the offline training process of the wind turbine transmission system fault early warning method of the present invention specifically includes the following steps:
s1: the method comprises the steps of obtaining SCADA operation data of a wind turbine generator data acquisition and monitoring control system, and selecting health state data of continuous operation of the wind turbine generator, and comprises the following specific steps:
s1.1: collecting data of a data acquisition and monitoring control system of the wind turbine generator set which continuously operates for 6 months, and screening out data of the wind turbine generator set in a normal operation state according to state codes in the data acquisition and monitoring control system;
s1.2: according to the time information of the wind turbine generator operation data, the continuity of the wind turbine generator operation data is checked, healthy and continuous wind turbine generator operation data is selected as a training set and a verification set, wherein 60% of the screened data is used as the training set, 20% of the screened data is used as a test set, and 20% of the screened data is used as the verification set.
S2: in order to reduce redundancy and feature dimension of data of a wind turbine generator data acquisition and monitoring control system, a correlation coefficient method is used for carrying out feature screening on the screened health state data of the wind turbine generator, and the specific method comprises the following steps:
s2.1: determining characteristic data related to the operation state of each component of a wind turbine transmission system, wherein temperature data are selected as key characteristic data by main components such as a gear box, a main bearing and a generator of the wind turbine transmission system, and the key characteristic data mainly comprise main bearing temperature, gear box oil temperature, gear box bearing temperature, gear box oil pump pressure, hydraulic system pressure, generator rotating speed, generator bearing temperature and the like;
s2.2: calculating linear correlation coefficient between variables by correlation coefficient method, and calculating linear correlation coefficient between two variables X and YIs calculated asWherein N is the number of samples, and X and Y are different variables; and (4) screening out the features with the correlation coefficient larger than 0.6 by taking the temperature features as a center, and finally determining that 25-dimensional input data exists.
S3: smoothing the selected data by adopting an Exponential Weighted Moving Average (EWMA) method, wherein a calculation formula is as follows, and the processed data is used as a training set of a model;
st=(1-ψ)st-1+ψdt
in the formula: t represents time; stFor the EWMA output value, s, corresponding to time tt-1For the EWMA output value corresponding to time t-1, dtFor the characteristic value of the data at the time t, psi represents the EWMA statistical weight of the historical data to the current data, and psi belongs to (0, 1)]In the experimental analysis, psi is taken to be 0.1 in the model.
S4: constructing a long-short term memory self-coding network model, training the model by using a training set, and completing data coding and reconstruction learning, as shown in fig. 2, the specific construction steps of the long-short term memory self-coding network model are as follows:
s4.1: and determining a training set, a testing set and a verification set, and determining the dimensions of the output layer and the input layer according to the screened data characteristics.
S4.2: and (3) training the model by using a single-layer neural network, and determining that the optimal number of the neural units of the first hidden layer is 128 by using a neural unit step-by-step increasing method.
S4.3: under the premise of determining the number of the first hidden layer nerve units, the depth of the network model is gradually increased in a mode of gradually decreasing the number of the nerve units layer by layer, and finally the structure of the model is determined to be 128-64-32-16-32-64-128.
S4.4: training the model by using an Adam algorithm, and determining parameters of the constructed long-short term memory self-coding network model;
s4.5: a denoising masking layer is added behind an input layer of the network model to enhance the anti-noise capability of the model and enable the model to have stronger robustness; the denoising mask layer represents the destruction degree of data, the original data represents the input model when the parameter is 0, the data representing the input model is completely destroyed when the parameter is 1, and finally the parameter of the Dropout layer is determined to be 0.2.
S4.6: and completing the construction of a network model, wherein the structure of the model is shown in FIG. 3, verifying the network through a verification set, judging whether false alarm occurs during normal work or not, judging whether early warning can be performed when a fault occurs or not, and storing the model after the model is confirmed to be effective.
S5: inputting the verification set into a trained long-short term memory self-coding network model, calculating a reconstruction residual error between reconstruction data and original input data, and setting a threshold control line by adopting a kernel density estimation method, wherein the specific method comprises the following steps:
s5.1: inputting the verification set Y into the trained long-short term memory self-coding network model, and calculating to obtain output reconstruction dataAnd calculating a reconstruction error E, the formula is as follows:
s5.2: the Robust Mahalanobis Distance (RMD) is adopted to complete the calculation of the monitoring index sequence h, and the kth sample EkIts corresponding monitoring index value hkThe calculation formula is as follows:
in the formula (I), the compound is shown in the specification,is a robust measure of central tendency, i.e., median; MCD-1Is an inverse covariance matrix calculated from a sample by a minimum covariance determination estimation method, and T represents transposition operation.
S5.3: determining a probability distribution function p (h) of a monitoring index sequence h by using a Kernel Density Estimation (KDE), wherein the calculation formula is as follows:
in the formula, σ is a kernel function broadband coefficient, K (·) is a kernel function, and N is the length of the monitoring index sequence h. Determining the value of sigma according to multiple experiments, and using Gaussian kernel as kernel functionWhere g is a function variable.
S5.4: and finally, calculating a corresponding monitoring threshold value d according to a probability density distribution function p (h) of the monitoring index sequence h obtained by estimation:
in the formula, α represents a confidence level and has a value of 0.99.
S6: and inputting data of the real-time operation of the wind turbine generator into the trained long-short term memory self-coding network model to calculate a reconstruction error, performing fault early warning when the reconstruction error is higher than a threshold value, and judging the specific component with the fault according to the size of the reconstruction error and the time when the reconstruction error reaches the threshold value. As shown in FIG. 4, the wind power transmission system fault early warning method based on the long-short term memory self-coding network model comprises the following steps:
firstly, relevant data obtained on line are input to a network model after being preprocessed, and reconstruction of the data is completed through model calculation.
And secondly, calculating a reconstruction error in real time according to the preprocessed data and the reconstructed data.
Again, the RMD values for each component are calculated separately from the reconstruction error.
Finally, comparing the RMD value of each part with a threshold value in real time, and if the RMD value of the part is larger than the threshold value, giving an early warning to the model and reporting the specific part; when each component RMD value is smaller than the threshold value, the normal operation of the transmission system of the unit is indicated. The method of the invention is used for testing the gearbox which is about to fail, and as shown in figure 5, the fault which is about to fail can be early warned 4.2 hours in advance.
According to the wind turbine generator transmission system fault early warning method based on the long-short term memory self-coding network, the wind turbine generator operation data is used, a sensor does not need to be installed secondarily, the characteristics of time-space relevance, nonlinearity and large noise of the data are considered, the occurrence of the generator fault can be effectively predicted in advance, the operation and maintenance cost can be well saved, and the productivity can be improved.
Claims (5)
1. A wind turbine generator transmission system fault early warning method is characterized by comprising the following steps:
step S1: acquiring running data of a wind turbine generator data acquisition and monitoring control system, and selecting health state data of continuous running of the wind turbine generator;
step S2: in order to reduce the redundancy and the feature dimension of the data of the wind turbine generator data acquisition and monitoring control system, a correlation coefficient method is used for carrying out feature screening on the screened health state data of the wind turbine generator;
step S3: smoothing the selected data by adopting an exponential weighted moving average method, and taking the processed data as a training set and a verification set of the model;
step S4: constructing a long-short term memory self-coding network model, and training the model by using a training set to complete data coding and reconstruction learning;
step S5: inputting the verification set into a trained long-short term memory self-coding network model, calculating a reconstruction residual error between reconstruction data and original input data, and setting a threshold control line by adopting a kernel density estimation method;
step S6: and inputting data of the real-time operation of the wind turbine generator into the trained long-short term memory self-coding network model to calculate a reconstruction error, performing fault early warning when the reconstruction error is higher than a threshold value, and judging the specific component with the fault according to the size of the reconstruction error and the time when the reconstruction error reaches the threshold value.
2. The wind turbine generator transmission system fault early warning method according to claim 1, wherein in step S1, the health state data is data of a data acquisition and monitoring control system acquired by the wind turbine generator in a normal operation state, and the specific step of selecting continuous health state data includes:
s1.1: collecting data of a data acquisition and monitoring control system of the wind turbine generator set which continuously operates for more than half a year, and screening out data of the wind turbine generator set in a normal operation state according to state codes in the data of the data acquisition and monitoring control system;
s1.2: and according to the time information of the wind turbine generator operation data, checking the continuity of the wind turbine generator operation data, and selecting healthy and continuous wind turbine generator operation data as a training set and a verification set.
3. The wind turbine generator transmission system fault early warning method according to claim 1, wherein the feature screening process in the step S2 specifically includes:
s2.1: determining characteristic data related to the operation state of each component of a wind turbine generator transmission system, wherein the characteristic data comprises main bearing temperature, gearbox oil temperature, gearbox bearing temperature, gearbox oil pump pressure, hydraulic system pressure, generator rotating speed and generator bearing temperature;
s2.2: the linear correlation coefficient between the variables is calculated by a correlation coefficient method, and the correlation coefficient r of the two variables X and Y is calculated asWherein N is the number of samples, and X and Y are different variables; and then selecting the first variables with larger values of r as the input of the long-short term memory self-coding model according to the size of the correlation coefficient r.
4. The wind turbine generator transmission system fault early warning method as claimed in claim 1, wherein the long-short term memory self-coding network model in the step S4 mainly comprises an input layer and a plurality of hidden layers and an output layer, the input layer and the output layer have the same dimension, wherein each hidden layer is composed of long-short term memory neural units, and the model is constructed by the following steps:
s4.1: determining a training set and a verification set, and determining the dimensions of an input layer and an output layer according to the screened data characteristics;
s4.2: training a model by using a single-layer neural network, and determining the optimal number of the first hidden layer neural units by a method of gradually increasing the neural units;
s4.3: under the premise of determining the number of the first hidden layer neural units, gradually increasing the depth of the network model in a mode of gradually decreasing the number of the neural units layer by layer;
s4.4: training the model by using an Adam algorithm, and determining parameters of the constructed long-short term memory self-coding network model;
s4.5: a denoising masking layer is added behind an input layer of the network model to enhance the anti-noise capability of the model and enable the model to have stronger robustness;
s4.6: and verifying the model by using the verification set, and saving the model when the model is determined to be capable of well reconstructing input data.
5. The wind turbine generator transmission system fault early warning method according to claim 1, wherein the step S5 of calculating a reconstruction error between original data and reconstructed data and calculating a threshold control line comprises the specific steps of:
s5.1: inputting the verification set Y into the trained long-short term memory self-coding network model, and calculating to obtain output reconstruction dataAnd calculating a reconstruction error E, the formula is as follows:
s5.2: calculating a monitoring index sequence h by using a Robust Mahalanobis Distance (RMD), and calculating a kth sample EkIts corresponding monitoring index value hkThe calculation formula is as follows:
in the formula (I), the compound is shown in the specification,is a robust measure of central tendency, i.e., median; MCD-1The method comprises the steps of determining an inverse covariance matrix obtained by calculation from a sample through a minimum covariance determination estimation method, wherein T represents transposition operation;
s5.3: determining a probability distribution function p (h) of a monitoring index sequence h by using a Kernel Density Estimation (KDE), wherein the calculation formula is as follows:
in the formula, σ is a kernel function broadband coefficient, K (·) is a kernel function, and N is the length of the monitoring index sequence h. Determining the value of sigma according to multiple experiments, and using Gaussian kernel as kernel functionWherein g is a function variable;
s5.4: and finally, calculating a corresponding monitoring threshold value d according to a probability density distribution function p (h) of the monitoring index sequence h obtained by estimation:
in the formula, α represents a confidence level.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011641131.XA CN112834211A (en) | 2020-12-31 | 2020-12-31 | Fault early warning method for transmission system of wind turbine generator |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011641131.XA CN112834211A (en) | 2020-12-31 | 2020-12-31 | Fault early warning method for transmission system of wind turbine generator |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112834211A true CN112834211A (en) | 2021-05-25 |
Family
ID=75926944
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011641131.XA Pending CN112834211A (en) | 2020-12-31 | 2020-12-31 | Fault early warning method for transmission system of wind turbine generator |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112834211A (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113408441A (en) * | 2021-06-24 | 2021-09-17 | 武汉理工大学 | Track bed fault early warning method based on DRSN and person correlation coefficient |
CN113700558A (en) * | 2021-09-01 | 2021-11-26 | 哈尔滨工业大学(威海) | Diesel engine air system fault detection method |
CN114048767A (en) * | 2021-10-28 | 2022-02-15 | 华电(福建)风电有限公司 | Fault monitoring and early warning method for wind power master control system |
CN114215706A (en) * | 2021-12-27 | 2022-03-22 | 南京邮电大学 | Wind turbine generator blade cracking fault early warning method and device |
CN114417704A (en) * | 2021-12-23 | 2022-04-29 | 中国大唐集团新能源科学技术研究院有限公司 | Wind turbine generator health assessment method based on improved stack type self-coding |
CN114813105A (en) * | 2022-04-11 | 2022-07-29 | 西安热工研究院有限公司 | Gear box fault early warning method and system based on working condition similarity evaluation |
CN114964370A (en) * | 2022-05-25 | 2022-08-30 | 国家电投集团科学技术研究院有限公司 | Wind turbine generator set frequency converter state monitoring method and system and electronic equipment |
CN115083123A (en) * | 2022-05-17 | 2022-09-20 | 中国矿业大学 | Mine coal spontaneous combustion intelligent grading early warning method taking measured data as drive |
CN116910570A (en) * | 2023-09-13 | 2023-10-20 | 华能新能源股份有限公司山西分公司 | Wind turbine generator system fault monitoring and early warning method and system based on big data |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109740727A (en) * | 2018-12-05 | 2019-05-10 | 中国软件与技术服务股份有限公司 | A kind of hydraulic turbine shaft state monitoring method neural network based and system |
CN110348513A (en) * | 2019-07-10 | 2019-10-18 | 北京华电天仁电力控制技术有限公司 | A kind of Wind turbines failure prediction method based on deep learning |
CN110610280A (en) * | 2018-10-31 | 2019-12-24 | 山东大学 | Short-term prediction method, model, device and system for power load |
US20200210824A1 (en) * | 2018-12-28 | 2020-07-02 | Utopus Insights, Inc. | Scalable system and method for forecasting wind turbine failure with varying lead time windows |
CN111415070A (en) * | 2020-03-03 | 2020-07-14 | 浙江运达风电股份有限公司 | Wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data |
CN111861272A (en) * | 2020-07-31 | 2020-10-30 | 西安交通大学 | Multi-source data-based complex electromechanical system abnormal state detection method |
-
2020
- 2020-12-31 CN CN202011641131.XA patent/CN112834211A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110610280A (en) * | 2018-10-31 | 2019-12-24 | 山东大学 | Short-term prediction method, model, device and system for power load |
CN109740727A (en) * | 2018-12-05 | 2019-05-10 | 中国软件与技术服务股份有限公司 | A kind of hydraulic turbine shaft state monitoring method neural network based and system |
US20200210824A1 (en) * | 2018-12-28 | 2020-07-02 | Utopus Insights, Inc. | Scalable system and method for forecasting wind turbine failure with varying lead time windows |
CN110348513A (en) * | 2019-07-10 | 2019-10-18 | 北京华电天仁电力控制技术有限公司 | A kind of Wind turbines failure prediction method based on deep learning |
CN111415070A (en) * | 2020-03-03 | 2020-07-14 | 浙江运达风电股份有限公司 | Wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data |
CN111861272A (en) * | 2020-07-31 | 2020-10-30 | 西安交通大学 | Multi-source data-based complex electromechanical system abnormal state detection method |
Non-Patent Citations (3)
Title |
---|
何群 等: "基于长短期记忆网络的风电机组齿轮箱故障预测", 《计量学报》 * |
王洪斌 等: "基于深度信念网络的风机主轴承状态监测方法", 《中国机械工程》 * |
陈俊生 等: "采用滑动窗口及多重加噪比堆栈降噪自编码的风电机组状态异常检测方法", 《电工技术学报》 * |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113408441A (en) * | 2021-06-24 | 2021-09-17 | 武汉理工大学 | Track bed fault early warning method based on DRSN and person correlation coefficient |
CN113408441B (en) * | 2021-06-24 | 2022-06-10 | 武汉理工大学 | Track bed fault early warning method based on DRSN and person correlation coefficient |
CN113700558A (en) * | 2021-09-01 | 2021-11-26 | 哈尔滨工业大学(威海) | Diesel engine air system fault detection method |
CN114048767A (en) * | 2021-10-28 | 2022-02-15 | 华电(福建)风电有限公司 | Fault monitoring and early warning method for wind power master control system |
CN114417704A (en) * | 2021-12-23 | 2022-04-29 | 中国大唐集团新能源科学技术研究院有限公司 | Wind turbine generator health assessment method based on improved stack type self-coding |
CN114215706A (en) * | 2021-12-27 | 2022-03-22 | 南京邮电大学 | Wind turbine generator blade cracking fault early warning method and device |
CN114215706B (en) * | 2021-12-27 | 2024-02-20 | 南京邮电大学 | Early warning method and device for cracking faults of wind turbine generator blades |
CN114813105A (en) * | 2022-04-11 | 2022-07-29 | 西安热工研究院有限公司 | Gear box fault early warning method and system based on working condition similarity evaluation |
CN115083123A (en) * | 2022-05-17 | 2022-09-20 | 中国矿业大学 | Mine coal spontaneous combustion intelligent grading early warning method taking measured data as drive |
CN114964370A (en) * | 2022-05-25 | 2022-08-30 | 国家电投集团科学技术研究院有限公司 | Wind turbine generator set frequency converter state monitoring method and system and electronic equipment |
CN116910570A (en) * | 2023-09-13 | 2023-10-20 | 华能新能源股份有限公司山西分公司 | Wind turbine generator system fault monitoring and early warning method and system based on big data |
CN116910570B (en) * | 2023-09-13 | 2023-12-15 | 华能新能源股份有限公司山西分公司 | Wind turbine generator system fault monitoring and early warning method and system based on big data |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112834211A (en) | Fault early warning method for transmission system of wind turbine generator | |
Fu et al. | Condition monitoring of wind turbine gearbox bearing based on deep learning model | |
US20230023931A1 (en) | Hydraulic turbine cavitation acoustic signal identification method based on big data machine learning | |
CN111537219B (en) | Fan gearbox performance detection and health assessment method based on temperature parameters | |
CN105550943A (en) | Method for identifying abnormity of state parameters of wind turbine generator based on fuzzy comprehensive evaluation | |
CN114841580A (en) | Generator fault detection method based on hybrid attention mechanism | |
CN115750228A (en) | Fault early warning method for pitch system of wind turbine generator | |
CN112836941A (en) | Online health condition evaluation method for high-pressure steam turbine system of coal-electric unit | |
CN113780375A (en) | Virtual-real interaction wind power plant wind power monitoring system based on digital twins and application | |
CN116609055A (en) | Method for diagnosing wind power gear box fault by using graph convolution neural network | |
Tang et al. | Prediction of bearing performance degradation with bottleneck feature based on LSTM network | |
Wang et al. | Wind turbine fault detection and identification through self-attention-based mechanism embedded with a multivariable query pattern | |
Yang et al. | Application research of arima model in wind turbine gearbox fault trend prediction | |
CN117034169A (en) | Power grid main transformer equipment abnormal state prediction method based on time sequence causality network | |
CN114320773B (en) | Wind turbine generator system fault early warning method based on power curve analysis and neural network | |
Song et al. | Anomaly detection of wind turbine generator based on temporal information | |
CN114841061A (en) | Wind power gear box health assessment method and system integrating timing sequence information | |
Hu et al. | Anomaly detection and remaining useful lifetime estimation based on degradation state for bearings | |
CN114417704A (en) | Wind turbine generator health assessment method based on improved stack type self-coding | |
CN115293189A (en) | Rotating machinery state monitoring method based on stack self-coding dimension reduction | |
Encalada-Dávila et al. | Wind turbine multi-fault detection based on SCADA data via an AutoEncoder | |
Ayman et al. | Fault Detection in Wind Turbines using Deep Learning | |
Zhou et al. | Offshore Wind Turbine Condition Assessment Method Based on GRU and Improved Fuzzy Comprehensive Evaluation | |
Liu et al. | Wind turbine spindle operating state recognition and early warning driven by SCADA data | |
Zeng et al. | Data-Driven Anomaly Detection Method Based on Similarities of Multiple Wind Turbines |
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 |