CN114320773B - Wind turbine generator system fault early warning method based on power curve analysis and neural network - Google Patents
Wind turbine generator system fault early warning method based on power curve analysis and neural network Download PDFInfo
- Publication number
- CN114320773B CN114320773B CN202111580061.6A CN202111580061A CN114320773B CN 114320773 B CN114320773 B CN 114320773B CN 202111580061 A CN202111580061 A CN 202111580061A CN 114320773 B CN114320773 B CN 114320773B
- Authority
- CN
- China
- Prior art keywords
- wind turbine
- early warning
- neural network
- prediction
- turbine generator
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Abstract
The invention discloses a wind turbine generator fault early warning method based on power curve analysis and a neural network, which fully utilizes SCADA data, does not need to analyze complex physical characteristics of a wind turbine generator, is respectively used for fault early warning of a pitch system and a yaw system of the wind turbine generator, further proves the effectiveness of the method, LPP feature extraction reduces modeling and prediction difficulty, improves prediction precision, and has the advantages that an extreme learning machine in a neural network algorithm has higher learning speed and generalization performance than a traditional BP neural network, is compared with an extreme learning machine prediction model, the stability and the prediction precision of the nuclear extreme learning machine prediction model are improved to a certain extent, and the change severity of the data can be quantified by combining the entropy method.
Description
Technical Field
The invention relates to the technical field of wind turbine automation, in particular to a wind turbine fault early warning method based on power curve analysis and a neural network.
Background
The wind turbine generator system has the advantages that the running environment of the wind turbine generator system is bad, the wind turbine generator system faults are frequently generated, the safety and the stability of the wind turbine generator system are brought to bad influence on the running, and serious challenges are brought to the development of wind power generation. Due to the variability and unpredictability of the running states of the wind turbine, the early warning research on the faults of the wind turbine has become an important research direction in wind power development. In recent years, with the development of artificial intelligence technology, more and more advanced intelligent algorithms are successfully applied to various fields, so that the method is selectively applied to the research of wind turbine generator fault early warning by adopting a neural network algorithm, and the predicted result is more accurate.
Disclosure of Invention
The invention aims to provide a wind turbine generator set fault early warning method based on power curve analysis and a neural network, which can solve the problems in the background technology.
In order to achieve the above purpose, the present invention provides the following technical solutions: a wind turbine generator system fault early warning method based on power curve analysis and a neural network comprises the following steps:
s101: data screening, namely screening SCADA data by using an algorithm combining least square and dispersion analysis according to a power characteristic curve so as to retain data conforming to normal working conditions of a unit, thereby improving the accuracy of a model;
s102: establishing a neural network model, screening input parameters of the model by using a random forest algorithm to simplify a model structure, and establishing the neural network model of the wind turbine generator by comparing the number of hidden layers;
s103: providing an index, and providing an index capable of reflecting the real-time running state of the unit by combining a sliding window model;
s104: determining a threshold value, and determining the threshold value of the index through a non-parameter estimation method;
s105: and (5) early warning and monitoring, and realizing state early warning and on-line monitoring.
The system comprises a working system and a management background, wherein the working system interacts data with the management background in a signal transmission mode, and the management background is provided with an application module for processing information.
Further, the working system comprises an analysis unit, a collection module, an early warning module and a transmission module, wherein the collection module is used for collecting state parameters including wind speed, wind direction, rotating speed, voltage, current, power, vibration, temperature and the like, and the collection module is connected with a management background through signals of the transmission module.
Further, the analysis unit comprises an angle module, a rotating speed module, a wind speed module and a power module, and utilizes relevant parameters of the SCADA system, such as rotating speed, power, wind speed, angle and the like, on the basis, a linear model for analyzing the electric quantity of each wind generating set is constructed by combining a generalized linear regression algorithm, and specific reasons for causing electric quantity loss are described.
Further, the working system further comprises a prediction judging module, wherein the prediction judging module can be used for full-field power prediction, blade icing prediction, mechanical state prediction of the wind generating set, standard comparison prediction of a single power curve, accuracy correction of wind speed prediction of a fan cabin and judgment of running health state.
Further, full-field power prediction refers to constructing a full-field power prediction model based on an ARIMA algorithm, using historical wind speed, wind direction, temperature, power and yaw data for training of the model, and using current corresponding data for prediction of the model.
Further, the mechanical state prediction of the wind turbine generator refers to deep excavation of parameters such as vibration signals, temperature, fan rotating speed and the like, a prediction model is established by utilizing a neural network, the model is trained by utilizing historical data, and then the mechanical state of the wind turbine generator is predicted by utilizing current parameters.
Further, for step S105, feature extraction may be performed on the state parameters of the wind turbine by using the office-preserving projection method, then a neural network prediction model of the target state parameters may be built, the residual variation trend of the target state parameter prediction model may be analyzed by using the information entropy method to determine whether the wind turbine is in a healthy state, if healthy, the wind turbine is allowed to perform normal operation and perform real-time monitoring on the wind turbine, if unhealthy, the early warning module may be started to perform early warning in time, and the early warning may be transmitted to the management background for preprocessing.
Compared with the prior art, the invention has the beneficial effects that: the method fully utilizes SCADA data, does not need to analyze complex physical characteristics of the wind turbine, is respectively used for fault early warning of a pitch system and a yaw system of the wind power plant, further proves the effectiveness of the method, LPP feature extraction reduces modeling and prediction difficulty, improves prediction accuracy, and has the advantages that an extreme learning machine in a neural network algorithm has higher learning speed and generalization performance than a traditional BP neural network.
Drawings
FIG. 1 is a flow chart summarizing a wind turbine generator system fault early warning method based on power curve analysis and a neural network;
FIG. 2 is a specific flowchart of a wind turbine generator system fault early warning method based on power curve analysis and a neural network;
FIG. 3 is a state early warning logic diagram of a wind turbine generator system fault early warning method based on power curve analysis and a neural network;
FIG. 4 is a system component block diagram of a wind turbine generator system fault early warning method based on power curve analysis and a neural network;
FIG. 5 is a block diagram of analysis unit components of the wind turbine generator fault early warning method based on power curve analysis and neural network;
FIG. 6 is a functional block diagram of a prediction judgment module of the wind turbine generator fault early warning method based on power curve analysis and a neural network.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1-2, a wind turbine generator system fault early warning method based on power curve analysis and a neural network comprises the following steps:
s101: data screening, namely screening SCADA data by using an algorithm combining least square and dispersion analysis according to a power characteristic curve so as to retain data conforming to normal working conditions of a unit, thereby improving the accuracy of a model;
s102: establishing a neural network model, screening input parameters of the model by using a random forest algorithm to simplify a model structure, and establishing the neural network model of the wind turbine generator by comparing the number of hidden layers;
s103: providing an index, and providing an index capable of reflecting the real-time running state of the unit by combining a sliding window model;
s104: determining a threshold value, and determining the threshold value of the index through a non-parameter estimation method;
s105: and (5) early warning and monitoring, and realizing state early warning and on-line monitoring.
Referring to fig. 4-5, a wind turbine generator system fault early warning method based on power curve analysis and neural network comprises a working system and a management background, wherein the working system interacts data with the management background in a signal transmission mode, the management background is provided with an application module for processing information, the working system comprises an analysis unit, an acquisition module, an early warning module and a transmission module, the acquisition module is used for acquiring state parameters including wind speed, wind direction, rotating speed, voltage, current, power, vibration, temperature and the like, the acquisition module is connected with the management background through the transmission module in a signal mode, the analysis unit comprises an angle module, a rotating speed module, a wind speed module and a power module, the analysis unit utilizes related parameters of a SCADA system such as rotating speed, power, wind speed, angle and the like, and a generalized linear regression algorithm is combined on the basis to construct a linear model which can be used for analyzing electric quantity of each wind turbine generator system, and specific reasons for causing electric quantity loss are explained.
Referring to fig. 6, the working system further includes a prediction and judgment module, wherein the prediction and judgment module can be used for full-field power prediction, blade icing prediction, wind turbine generator mechanical state prediction, single machine power curve benchmarking prediction, fan cabin wind speed prediction accuracy correction and judgment of running health state, the full-field power prediction is based on an ARIMA algorithm to construct a full-field power prediction model, historical wind speed, wind direction, temperature, power and yaw data are used for model training, current corresponding data are used for model prediction, wind turbine generator mechanical state prediction is parameters such as deep excavation vibration signals, temperature and fan rotating speed, the neural network is used for building a prediction model, the model is trained, and current parameters are used for wind turbine generator mechanical state prediction.
Early warning working principle: the method comprises the steps of firstly extracting the characteristics of the state parameters of the wind turbine by using a local protection projection method, then establishing a neural network prediction model of the target state parameters, analyzing the residual variation trend of the target state parameter prediction model by using an information entropy method to judge whether the wind turbine is in a healthy state, if so, allowing the wind turbine to normally work and monitoring the wind turbine in real time, and if not, starting an early warning module to timely warn and transmitting the early warning module to a management background for preprocessing.
The foregoing is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art, who is within the scope of the present invention, should be covered by the protection scope of the present invention by making equivalents and modifications to the technical solution and the inventive concept thereof.
Claims (8)
1. A wind turbine generator system fault early warning method based on power curve analysis and a neural network is characterized by comprising the following steps:
s101: data screening, namely screening SCADA data by using an algorithm combining least square and dispersion analysis according to a power characteristic curve so as to retain data conforming to normal working conditions of a unit, thereby improving the accuracy of a model;
s102: establishing a neural network model, screening input parameters of the model by using a random forest algorithm to simplify a model structure, and establishing the neural network model of the wind turbine generator by comparing the number of hidden layers;
s103: providing an index, and providing an index capable of reflecting the real-time running state of the unit by combining a sliding window model;
s104: determining a threshold value, and determining the threshold value of the index through a non-parameter estimation method;
s105: and (5) early warning and monitoring, and realizing state early warning and on-line monitoring.
2. The wind turbine generator system fault early warning method based on power curve analysis and neural network according to claim 1, wherein the method comprises the following steps: the system comprises a working system and a management background, wherein the working system interacts data with the management background in a signal transmission mode, and the management background is provided with an application module for processing information.
3. The wind turbine generator system fault early warning method based on power curve analysis and neural network as claimed in claim 2, wherein the method is characterized in that: the working system comprises an analysis unit, a collection module, an early warning module and a transmission module, wherein the collection module is used for collecting state parameters including wind speed, wind direction, rotating speed, voltage, current, power, vibration, temperature and the like, and the collection module is connected with a management background through signals of the transmission module.
4. The wind turbine generator system fault early warning method based on power curve analysis and neural network as claimed in claim 3, wherein the method is characterized by comprising the following steps of: the analysis unit comprises an angle module, a rotating speed module, a wind speed module and a power module, and the analysis unit utilizes relevant parameters of the SCADA system to construct a linear model for analyzing the electric quantity of each wind generating set by combining a generalized linear regression algorithm on the basis, so that specific reasons for causing electric quantity loss are explained.
5. The wind turbine generator system fault early warning method based on power curve analysis and neural network as claimed in claim 2, wherein the method is characterized in that: the working system further comprises a prediction judging module, wherein the prediction judging module can be used for full-field power prediction, blade icing prediction, mechanical state prediction of the wind generating set, standard comparison prediction of a single power curve, accuracy correction of wind speed prediction of a fan cabin and judgment of running health states.
6. The wind turbine generator system fault early warning method based on power curve analysis and neural network according to claim 5, wherein the method comprises the following steps: full-field power prediction is to construct a full-field power prediction model based on ARIMA algorithm, use historical wind speed, wind direction, temperature, power and yaw data for training of the model, and use current corresponding data for prediction of the model.
7. The wind turbine generator system fault early warning method based on power curve analysis and neural network according to claim 5, wherein the method comprises the following steps: the wind turbine generator set mechanical state prediction refers to deep excavation of vibration signal, temperature and fan rotating speed parameters, a prediction model is established by utilizing a neural network, the model is trained by utilizing historical data, and then the current parameters are utilized to predict the wind turbine generator set mechanical state.
8. The wind turbine generator system fault early warning method based on power curve analysis and neural network according to claim 1, wherein the method comprises the following steps: for step S105, feature extraction may be performed on the state parameters of the wind turbine by using the office-preserving projection method, then a neural network prediction model of the target state parameters is established, the residual variation trend of the target state parameter prediction model is analyzed by using the information entropy method to determine whether the wind turbine is in a healthy state, if so, the wind turbine is allowed to work normally and monitor the wind turbine in real time, and if not, an early warning module is started to perform early warning in time and transmitted to the management background for preprocessing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111580061.6A CN114320773B (en) | 2021-12-22 | 2021-12-22 | Wind turbine generator system fault early warning method based on power curve analysis and neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111580061.6A CN114320773B (en) | 2021-12-22 | 2021-12-22 | Wind turbine generator system fault early warning method based on power curve analysis and neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114320773A CN114320773A (en) | 2022-04-12 |
| CN114320773B true CN114320773B (en) | 2023-09-22 |
Family
ID=81055281
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111580061.6A Active CN114320773B (en) | 2021-12-22 | 2021-12-22 | Wind turbine generator system fault early warning method based on power curve analysis and neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114320773B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115079626B (en) * | 2022-07-21 | 2022-11-22 | 东方电气风电股份有限公司 | Early warning method and system for potential operation risk of wind generating set component |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016086360A1 (en) * | 2014-12-02 | 2016-06-09 | Abb Technology Ltd | Wind farm condition monitoring method and system |
| WO2017113998A1 (en) * | 2015-12-31 | 2017-07-06 | 北京金风科创风电设备有限公司 | Computer storage medium, computer program product, and method and device for monitoring for malfunction of a wind turbine |
| KR102031121B1 (en) * | 2018-06-05 | 2019-10-11 | 한남대학교 산학협력단 | Monitoring system of wind generator using radar |
| CN112648140A (en) * | 2020-12-21 | 2021-04-13 | 北京华能新锐控制技术有限公司 | Fault tolerance method for wind turbine generator pitch angle encoder based on signal reconstruction |
| CN112733283A (en) * | 2020-12-21 | 2021-04-30 | 北京华能新锐控制技术有限公司 | Wind turbine generator component fault prediction method |
-
2021
- 2021-12-22 CN CN202111580061.6A patent/CN114320773B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016086360A1 (en) * | 2014-12-02 | 2016-06-09 | Abb Technology Ltd | Wind farm condition monitoring method and system |
| WO2017113998A1 (en) * | 2015-12-31 | 2017-07-06 | 北京金风科创风电设备有限公司 | Computer storage medium, computer program product, and method and device for monitoring for malfunction of a wind turbine |
| KR102031121B1 (en) * | 2018-06-05 | 2019-10-11 | 한남대학교 산학협력단 | Monitoring system of wind generator using radar |
| CN112648140A (en) * | 2020-12-21 | 2021-04-13 | 北京华能新锐控制技术有限公司 | Fault tolerance method for wind turbine generator pitch angle encoder based on signal reconstruction |
| CN112733283A (en) * | 2020-12-21 | 2021-04-30 | 北京华能新锐控制技术有限公司 | Wind turbine generator component fault prediction method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114320773A (en) | 2022-04-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Wind turbine fault detection based on SCADA data analysis using ANN | |
| CN106662072B (en) | Wind-driven generator method for monitoring state and system | |
| CN102520697B (en) | Onsite information preprocessing method of remote cooperative diagnosis | |
| CN109543210A (en) | A kind of Wind turbines failure prediction system based on machine learning algorithm platform | |
| CN105205569B (en) | State of fan gear box online evaluation method for establishing model and online evaluation method | |
| CN103150473A (en) | Method and device for monitoring and diagnosing generating efficiency of wind turbine generator in real-time manner | |
| CN111582392B (en) | Multi-working-condition health state online monitoring method for key components of wind turbine generator | |
| CN105550943A (en) | Method for identifying abnormity of state parameters of wind turbine generator based on fuzzy comprehensive evaluation | |
| CN113204921B (en) | Method and system for predicting remaining service life of airplane turbofan engine | |
| CN112834211A (en) | Fault early warning method for transmission system of wind turbine generator | |
| CN113339204A (en) | Wind driven generator fault identification method based on hybrid neural network | |
| CN114320773B (en) | Wind turbine generator system fault early warning method based on power curve analysis and neural network | |
| CN114519923A (en) | Intelligent diagnosis and early warning method and system for power plant | |
| Li et al. | Application of Big Data Technology in Evaluation of Operating State of High-Pressure Hydrogenation Heat Exchanger | |
| CN116779202A (en) | Digital twinning-based intelligent fault diagnosis method and system for rotating equipment of nuclear power plant | |
| Zhang et al. | A novel fault diagnosis method for wind turbine based on adaptive multivariate time-series convolutional network using SCADA data | |
| CN114169718A (en) | Method for improving reliability of wind turbine generator based on state evaluation of wind turbine generator | |
| Zhang | Comparison of data-driven and model-based methodologies of wind turbine fault detection with SCADA data | |
| Souza et al. | Evaluation of data based normal behavior models for fault detection in wind turbines | |
| Song et al. | Anomaly detection of wind turbine generator based on temporal information | |
| Fan et al. | An intelligent condition-based monitoring and maintenance system for wind turbine | |
| Zhou et al. | Offshore Wind Turbine Condition Assessment Method Based on GRU and Improved Fuzzy Comprehensive Evaluation | |
| CN117034157B (en) | Hydropower equipment fault identification method and system combining multimodal operation data | |
| Ou et al. | Fault Prediction Model of Wind Power Pitch System Based on BP Neural Network | |
| Liu et al. | Wind Turbine Fault Diagnosis Based on Feature Selection and Stacking Model Fusion with Small-scale Data. |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20221014 Address after: No. 2, xizhuan Hutong, Caishikou street, Xicheng District, Beijing 100052 Applicant after: Datang renewable energy test and Research Institute Co.,Ltd. Address before: 100041 room a-1725, 2nd floor, building 3, yard 30, Shixing street, Shijingshan District, Beijing Applicant before: CHINA DATANG CORPORATION NEW ENERGY SCIENCE AND TECHNOLOGY RESEARCH INSTITUTE Co.,Ltd. |
|
| TA01 | Transfer of patent application right | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |