CN115095487A - Wind turbine state monitoring method based on multi-source heterogeneous SCADA data - Google Patents
Wind turbine state monitoring method based on multi-source heterogeneous SCADA data Download PDFInfo
- Publication number
- CN115095487A CN115095487A CN202210778796.8A CN202210778796A CN115095487A CN 115095487 A CN115095487 A CN 115095487A CN 202210778796 A CN202210778796 A CN 202210778796A CN 115095487 A CN115095487 A CN 115095487A
- Authority
- CN
- China
- Prior art keywords
- wind turbine
- average
- monitoring
- turbine generator
- wind
- 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
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F03—MACHINES OR ENGINES FOR LIQUIDS; WIND, SPRING, OR WEIGHT MOTORS; PRODUCING MECHANICAL POWER OR A REACTIVE PROPULSIVE THRUST, NOT OTHERWISE PROVIDED FOR
- F03D—WIND MOTORS
- F03D17/00—Monitoring or testing of wind motors, e.g. diagnostics
-
- 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
-
- 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
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E10/00—Energy generation through renewable energy sources
- Y02E10/70—Wind energy
- Y02E10/72—Wind turbines with rotation axis in wind direction
Landscapes
- Engineering & Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Biophysics (AREA)
- Data Mining & Analysis (AREA)
- Combustion & Propulsion (AREA)
- Chemical & Material Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Sustainable Energy (AREA)
- Computational Linguistics (AREA)
- Mechanical Engineering (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Sustainable Development (AREA)
- Wind Motors (AREA)
Abstract
The invention provides a wind turbine state monitoring method based on multi-source heterogeneous SCADA data, which comprises the following steps: calculating the probability distribution of the monitoring quantities of all wind turbine generators in the wind field, and screening out one wind turbine generator which can represent the whole wind field; cleaning monitoring data; performing feature dimensionality reduction, and screening out a plurality of sensitive feature parameters with high correlation degree with the monitored quantity; normalizing the screened sensitive characteristic parameters, then establishing a sample set, and training a long-term memory neural network model by using the sample set; calculating the root mean square error between the real value and the predicted value of the monitored quantity at the current moment, and constructing a health monitoring state index of the wind turbine generator; and carrying out sliding average by designing a sliding window; when the screened wind turbine generators are monitored, when the health monitoring state index exceeds the early warning threshold value, an alarm is given out. The invention realizes the state monitoring and the fault early warning of the wind turbine generator.
Description
Technical Field
The invention belongs to the field of intelligent manufacturing, and particularly relates to a wind turbine state monitoring method based on multi-source heterogeneous SCADA data.
Background
In recent years, the world has increasingly paid attention to the problems of energy safety, ecological environment, climate change and the like, and the accelerated development of the wind power industry has become a common consensus and consistent action for promoting energy transformation and coping with global climate change in the international society. The wind turbine generator set runs in extreme severe weather such as insolation, sand and dust, rain and snow, thunderstorm, high temperature or severe cold for a long time, so that the failure of the wind turbine generator set frequently occurs, the performance is continuously reduced, and a great amount of economic loss is caused. The research and development of the intelligent state monitoring technology of the wind turbine generator are urgent for the national strategic demands.
However, the existing wind turbine generator set vibration data is high in sampling frequency, long in acquisition time and multiple in measuring points, vibration monitoring big data is formed, difficulty is brought to remote wireless transmission and storage of data, and a monitoring model of a single unit is difficult to popularize in other units. By utilizing the characteristics of small volume, easy storage and transmission of multi-source heterogeneous SCADA data, a wind turbine state monitoring method based on the multi-source heterogeneous SCADA data is needed to be developed.
Disclosure of Invention
Aiming at the problems, the invention provides a wind turbine state monitoring method based on multi-source heterogeneous SCADA data, which is used for constructing a long-time memory neural network prediction model, fully mining the time sequence information of characteristic parameters and the coupling information between the characteristic parameters and realizing the state monitoring and the fault early warning of the wind turbine. In order to achieve the technical purpose, the embodiment of the invention adopts the technical scheme that:
the embodiment of the invention provides a wind turbine state monitoring method based on multi-source heterogeneous SCADA data, which comprises the following steps:
step S10, acquiring monitoring data of the wind turbine generator as a characteristic parameter, and selecting a characteristic parameter capable of representing the running state of the wind turbine generator as a monitoring quantity; calculating the probability distribution of the monitoring quantities of all the wind turbines in the wind field, and screening out one wind turbine which can represent the whole wind field based on the monitoring quantity probability distribution similarity;
step S20, cleaning the monitoring data, and eliminating invalid data in the monitoring data of the wind turbine generator;
step S30, performing feature dimension reduction on all feature parameters of the wind turbine generator based on the Spearman correlation coefficient, and screening out a plurality of sensitive feature parameters with high correlation degree with the monitored quantity;
step S40, normalizing the screened sensitive characteristic parameters, then establishing a sample set, and training a long-term memory neural network model by using the sample set; the sample set comprises a training set and a test set;
step S50, for one screened wind turbine generator which can represent the whole wind field, inputting the data concentrated by testing into a long-time memory neural network model which completes training, and obtaining a predicted value of the monitoring quantity at the current moment; calculating the root mean square error between the true value and the predicted value of the monitoring quantity at the current moment, and constructing a health monitoring state index HI (t) of the wind turbine generator d ) (ii) a And carrying out sliding average by designing a sliding window;
step S60, setting an early warning threshold value, and when monitoring the screened wind turbine generator, judging the health monitoring state index HI (t) of the wind turbine generator d ) And when the early warning threshold value is exceeded, an alarm is given out.
Further, in step S20, the conditions for monitoring data cleansing are as follows:
firstly, the average wind speed is less than or equal to the cut-in wind speed; or the like, or a combination thereof,
the average wind speed is larger than or equal to the cut-out wind speed; or the like, or, alternatively,
③ the average main shaft rotating speed is less than or equal to 10 rpm.
Further, the monitored quantity is an average gearbox oil temperature;
the screened sensitive characteristic parameters comprise: average generator speed, average nacelle temperature, average gearbox low speed end bearing temperature, average gearbox oil filter inlet pressure, average gearbox oil distributor outlet pressure, average gearbox oil temperature at first 1 moment, average gearbox oil temperature at first 2 moments.
Further, the hyper-parameters of the long-term memory neural network model comprise:
the method comprises the steps of inputting the number of features, the number of neurons in each layer, time step length, the number of output features, learning rate and training times.
Further, the health monitoring state index HI (t) of the wind turbine generator d ) As follows:
wherein, t d Is the current time, r real Representing the real value of the average gearbox oil temperature; r is predict Representing an average gearbox oil temperature predicted value; l represents the sliding window length.
Further, the setting of the early warning threshold specifically includes:
calculating a health monitoring State indicator HI (t) d ) The Gaussian kernel density distribution takes a Gaussian kernel density function value with the confidence probability of 99.95 percent as an early warning threshold value.
The technical scheme provided by the embodiment of the invention has the following beneficial effects:
1) the automation of the state monitoring of the wind turbine generator is realized.
2) The adaptability is relatively good, and the long-time memory neural network model can be suitable for almost all wind turbines in a wind power plant.
3) The health monitoring state index has accurate indication, and the real-time monitoring of the running health state of the wind turbine generator is realized.
Drawings
Fig. 1 is a flowchart of a monitoring method according to an embodiment of the present invention.
Fig. 2 is a schematic diagram of index change of a health monitoring state of a normal wind turbine generator in the embodiment of the present invention.
Fig. 3 is a schematic diagram of index change of a health monitoring state of a faulty wind turbine in an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
As shown in fig. 1, a wind turbine state monitoring method based on multisource heterogeneous SCADA data provided by an embodiment of the present invention includes the following steps:
step S10, acquiring monitoring data of the wind turbine generator as a characteristic parameter, and selecting a characteristic parameter capable of representing the running state of the wind turbine generator as a monitoring quantity; calculating the probability distribution of the monitoring quantities of all the wind turbines in the wind field, and screening out one wind turbine which can represent the whole wind field based on the monitoring quantity probability distribution similarity;
in the embodiment, the monitored quantity is an average gearbox oil temperature;
step S20, cleaning the monitoring data and eliminating invalid data in the monitoring data of the wind turbine generator;
in this embodiment, the conditions for monitoring data cleansing are as follows:
firstly, the average wind speed is less than or equal to 3m/s (cut-in wind speed); or the like, or, alternatively,
the average wind speed is more than or equal to 25m/s (cut-out wind speed); or the like, or, alternatively,
thirdly, the average main shaft rotating speed is less than or equal to 10 rpm;
step S30, performing feature dimensionality reduction on all feature parameters of the wind turbine generator based on the Spearman correlation coefficient, and screening out a plurality of sensitive feature parameters with high correlation degree with the monitored quantity;
in this embodiment, the screened sensitive characteristic parameters include: average generator speed, average nacelle temperature, average gearbox low speed end bearing temperature, average gearbox oil filter inlet pressure, average gearbox oil distributor outlet pressure, average gearbox oil temperature at first 1 moment, average gearbox oil temperature at first 2 moments; seven characteristic parameters in total;
it should be noted that the monitoring data of the wind turbine generator is sampled at certain time intervals, so that the monitoring data correspond to the current time, the previous 1 time and the previous 2 times;
step S40, normalizing the screened sensitive characteristic parameters, then establishing a sample set, and training a long-term memory neural network model by using the sample set;
the numerical ranges of the 7 screened sensitive characteristic parameters are as follows:
maximum value | Minimum value of | Unit of | |
Average generator speed | 2000 | 1000 | rpm |
Average cabin temperature | 30 | -30 | ℃ |
Averaging gearbox low speed end bearing temperature | 100 | 0 | ℃ |
Averaging gearbox oil filter inlet pressure | 10 | 0 | bar |
Averaging gearbox oil distributor outlet pressure | 10 | 0 | bar |
Average gearbox oil temperature 1 | 100 | 0 | ℃ |
Average gearbox oil temperature 2 | 100 | 0 | ℃ |
The normalization formula of the sensitive characteristic parameters is as follows:
wherein X is a sensitive characteristic parameter value, X min 、x max Respectively the minimum value and the maximum value of the sensitive characteristic parameter;
80& in the sample set are used as training sets, and 20% are used as test sets;
the hyper-parameters of the long-time memory neural network model comprise:
the structure of a neural network model is memorized at long time and short time: an input layer +3 hidden layers + an output layer;
inputting characteristic quantities, namely average generator rotating speed, average cabin temperature, average gearbox low-speed end bearing temperature, average gearbox oil filter inlet pressure, average gearbox oil distributor outlet pressure, average gearbox oil temperature at the previous 1 moment and average gearbox oil temperature at the previous 2 moments; 7 in total;
the number of neurons per layer was 128;
the time step is 1; the memory duration of the long-and-short memory neural network model is represented, and the longer the time step length is, the higher the risk of gradient disappearance or explosion is;
outputting characteristic quantity, namely average gearbox oil temperature at the current moment;
step S50, for one wind turbine generator which can represent the whole wind field is screened out, the data in the test set is input into a neural network model which is memorized at long time and finishing training, and the predicted value of the monitoring quantity at the current moment is obtained; calculating the root mean square error between the true value and the predicted value of the monitoring quantity at the current moment, and constructing a health monitoring state index HI (t) of the wind turbine generator d ) (ii) a And carrying out sliding average by designing a sliding window; index burrs can be reduced, the trend is improved, and the real-time monitoring on the running health state of the wind turbine generator is realized;
in the embodiment, the monitored quantity is the average gearbox oil temperature;
health monitoring state index HI (t) of wind turbine generator d ) As follows:
wherein, t d Is the current time, r real Representing the real value of the average gearbox oil temperature; r is predict Representing an average gearbox oil temperature predicted value; l represents the sliding window length;
step S60, setting an early warning threshold value, and when monitoring the screened wind turbine generator, judging the health monitoring state index HI (t) of the wind turbine generator d ) When the early warning threshold value is exceeded, an alarm is sent out;
the setting of the early warning threshold specifically comprises:
calculating a health monitoring State indicator HI (t) d ) The gaussian kernel density distribution takes a gaussian kernel density function value with a confidence probability of 99.95% as an early warning threshold, which is 0.025 in this embodiment;
in a specific embodiment, the method provided by the application is verified by using SCADA data of a wind turbine generator of a certain wind power plant; the rated power of the wind turbine of the wind power plant is 2MW, the sampling interval of SCADA data is 5min, and the time span of data selection is 2020, 6 and 1 months, 2021, 7 and 27 months; each SCADA data comprises 100 characteristic parameters;
calculating the probability distribution of the average gearbox oil temperatures of all the wind turbines, calculating the similarity of the probability distribution, and selecting one wind turbine which can represent the whole wind field; the wind turbine generator set and the average gear box of other generator sets have higher oil temperature probability distribution similarity; in the test and analysis, the No. 004 wind turbine generator is screened out;
according to the probability distribution diagram of the health monitoring state indexes of the No. 004 wind turbine generator, when the confidence coefficient is 99.95%, the determined early warning threshold value is 0.025, as can be seen from the graph in FIG. 2, the health monitoring state indexes of the wind turbine generator do not exceed the early warning threshold value of 0.025, the whole is stable, no obvious decline trend exists, and the fact that the No. 004 wind turbine generator is normal and consistent with an actual result is shown;
for a fault wind turbine generator (No. 002 wind turbine generator), as shown in fig. 3, it can be seen that a health monitoring state index has an obvious degradation trend from a certain day, and the health monitoring state index exceeds an early warning threshold value of 0.025, which indicates that No. 002 wind turbine generator has a fault, and is consistent with an actual result.
For other wind turbines in the wind power plant, only the real-time monitoring data of the other wind turbines are needed to fine-tune the hyper-parameters of the trained long-time and short-time memory neural network prediction model, and the long-time and short-time memory neural network prediction model is also suitable for the other wind turbines. The state monitoring of other wind turbines is the same as above.
Finally, it should be noted that the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting, and although the present invention has been described in detail with reference to examples, it should be understood by those skilled in the art that modifications or equivalent substitutions may be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, which should be covered by the claims of the present invention.
Claims (6)
1. A wind turbine state monitoring method based on multi-source heterogeneous SCADA data is characterized by comprising the following steps:
step S10, acquiring monitoring data of the wind turbine generator as a characteristic parameter, and selecting a characteristic parameter capable of representing the running state of the wind turbine generator as a monitoring quantity; calculating the probability distribution of the monitoring quantities of all the wind turbines in the wind field, and screening out one wind turbine which can represent the whole wind field based on the monitoring quantity probability distribution similarity;
step S20, cleaning the monitoring data and eliminating invalid data in the monitoring data of the wind turbine generator;
step S30, performing feature dimensionality reduction on all feature parameters of the wind turbine generator based on the Spearman correlation coefficient, and screening out a plurality of sensitive feature parameters with high correlation degree with the monitored quantity;
step S40, normalizing the screened sensitive characteristic parameters, then establishing a sample set, and training a long-term memory neural network model by using the sample set; the sample set comprises a training set and a test set;
step S50, for one wind turbine generator which can represent the whole wind field is screened out, the data in the test set is input into a neural network model which is memorized at long time and finishing training, and the predicted value of the monitoring quantity at the current moment is obtained; calculating the root mean square error between the true value and the predicted value of the monitoring quantity at the current moment, and constructing a health monitoring state index HI (t) of the wind turbine generator d ) (ii) a And carrying out sliding average by designing a sliding window;
step S60, setting an early warning threshold value, and when monitoring the screened wind turbine generator, judging the health monitoring state index HI (t) of the wind turbine generator d ) And when the early warning threshold value is exceeded, an alarm is given out.
2. The wind turbine condition monitoring method based on multisource heterogeneous SCADA data of claim 1,
in step S20, the conditions for monitoring data cleansing are as follows:
firstly, the average wind speed is less than or equal to the cut-in wind speed; or the like, or, alternatively,
the average wind speed is larger than or equal to the cut-out wind speed; or the like, or, alternatively,
and thirdly, the average main shaft rotating speed is less than or equal to 10 rpm.
3. The wind turbine condition monitoring method based on multisource heterogeneous SCADA data of claim 1,
the monitored quantity is the average gearbox oil temperature;
the screened sensitive characteristic parameters comprise: average generator speed, average nacelle temperature, average gearbox low speed end bearing temperature, average gearbox oil filter inlet pressure, average gearbox oil distributor outlet pressure, average gearbox oil temperature at first 1 moment, average gearbox oil temperature at first 2 moments.
4. The wind turbine condition monitoring method based on multisource heterogeneous SCADA data of claim 3,
the hyper-parameters of the long-time memory neural network model comprise:
the method comprises the steps of inputting the number of features, the number of neurons in each layer, time step length, the number of output features, learning rate and training times.
5. The wind turbine condition monitoring method based on multisource heterogeneous SCADA data of claim 3,
health monitoring state index HI (t) of wind turbine generator d ) As follows:
wherein, t d Is the current time, r real Indicating average gearbox oil temperature trueReal value; r is predict Representing an average gearbox oil temperature predicted value; l represents the sliding window length.
6. The wind turbine condition monitoring method based on multi-source heterogeneous SCADA data according to any one of claims 1 to 5,
the setting of the early warning threshold specifically comprises:
calculating a health monitoring State indicator HI (t) d ) The Gaussian kernel density distribution takes a Gaussian kernel density function value with the confidence probability of 99.95 percent as an early warning threshold value.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210778796.8A CN115095487A (en) | 2022-06-30 | 2022-06-30 | Wind turbine state monitoring method based on multi-source heterogeneous SCADA data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210778796.8A CN115095487A (en) | 2022-06-30 | 2022-06-30 | Wind turbine state monitoring method based on multi-source heterogeneous SCADA data |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115095487A true CN115095487A (en) | 2022-09-23 |
Family
ID=83295243
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210778796.8A Pending CN115095487A (en) | 2022-06-30 | 2022-06-30 | Wind turbine state monitoring method based on multi-source heterogeneous SCADA data |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115095487A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116756644A (en) * | 2023-08-16 | 2023-09-15 | 华北电力大学 | Early warning method for icing fault of anemoclinograph of wind turbine generator |
-
2022
- 2022-06-30 CN CN202210778796.8A patent/CN115095487A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116756644A (en) * | 2023-08-16 | 2023-09-15 | 华北电力大学 | Early warning method for icing fault of anemoclinograph of wind turbine generator |
CN116756644B (en) * | 2023-08-16 | 2023-11-21 | 华北电力大学 | Early warning method for icing fault of anemoclinograph of wind turbine generator |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Rehman | Long-term wind speed analysis and detection of its trends using Mann–Kendall test and linear regression method | |
CN105550943A (en) | Method for identifying abnormity of state parameters of wind turbine generator based on fuzzy comprehensive evaluation | |
CN110544003B (en) | Wind turbine generator system frequency converter state evaluation method based on temperature prediction | |
CN111415070A (en) | Wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data | |
Dongmei et al. | Research on wind power forecasting in wind farms | |
CN115095487A (en) | Wind turbine state monitoring method based on multi-source heterogeneous SCADA data | |
CN113516271A (en) | Wind power cluster power day-ahead prediction method based on space-time neural network | |
CN116609055A (en) | Method for diagnosing wind power gear box fault by using graph convolution neural network | |
CN111950764A (en) | Extreme weather condition power grid wind power prediction correction method | |
Wilson et al. | Modeling the relationship between wind turbine failure modes and the environment | |
CN117365677A (en) | Steam turbine unit performance health state evaluation method | |
CN115204712B (en) | Offshore and coastal wind power plant site selection evaluation method | |
CN113111314B (en) | Multi-unit cooperation based health state evaluation method for grid-connected wind turbine generator | |
CN113761692B (en) | Method for identifying running states of multiple wind turbine generators based on migration component analysis | |
CN110334951B (en) | Intelligent evaluation method and system for high-temperature capacity reduction state of wind turbine generator | |
CN114607571A (en) | Offshore wind power gear box fault identification method and system based on lubricating system monitoring | |
CN114048767A (en) | Fault monitoring and early warning method for wind power master control system | |
CN113847214A (en) | Icing detection method for large wind generating set | |
CN113822344B (en) | Method for monitoring state of front bearing of generator of wind turbine generator based on data driving | |
Liu et al. | Failure Probability Model of Distribution Network Equipment Based on Improved Age-Reduction Model and Factor Correction | |
Xie et al. | Ultra-short-term Forecasting Method of Photovoltaic Power Generation Based on Combined Deep Neural Network | |
Ghosh | Wind speed Weibull distribution and wind energy potential of Chandpur, Bangladesh | |
CN112488417B (en) | Power grid operation characteristic sensing method and system based on LBP and neural network | |
CN118036347B (en) | Wind power generation full-flow error tracing method and system suitable for extreme weather | |
CN118035875A (en) | Wind power generation power prediction method and system based on extreme weather influence factors |
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 |