A big data processing method for fan safety monitoring
By integrating GNSS and IMU data for monitoring, and combining Kalman filtering and BP neural network big data analysis methods, the problem of insufficient data analysis in the wind turbine safety monitoring system has been solved, enabling accurate prediction and early warning of wind turbine displacement, and improving the stability of wind turbine safe operation and maintenance efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HUNAN UNIV
- Filing Date
- 2023-10-25
- Publication Date
- 2026-06-19
AI Technical Summary
Existing wind turbine safety monitoring systems cannot fully analyze massive amounts of data, nor can they model and predict wind turbine deformation, thus failing to provide effective suggestions and status predictions for the safe operation of wind turbines.
The system employs GNSS and IMU data fusion monitoring, utilizes Kalman filtering and BP neural networks for data processing, and combines big data analysis methods to dynamically adjust the Kalman gain, identify abnormal data, and issue early warnings.
It improves the accuracy and stability of wind turbine monitoring, can extract wind turbine displacement trend characteristics, make predictions and provide early warnings, dynamically adjust the Kalman filter gain, identify abnormal wind turbines, and perform timely maintenance.
Smart Images
Figure CN117249051B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of wind turbine safety monitoring technology, specifically to a big data processing method for wind turbine safety monitoring. Background Technology
[0002] With the construction of wind turbines and the continuous increase in the capacity of individual units, the maintenance and safety monitoring of wind turbines has become a challenge.
[0003] Currently, wind turbine safety monitoring primarily utilizes sensors for wind force, vibration, and tilt detection. Vibration sensors monitor and analyze the turbine's vibration signals to obtain vibration parameters such as acceleration, velocity, and displacement. Analyzing collected wind force data allows for the shutdown of wind turbines in high wind conditions. Analysis of tilt sensor data yields parameters such as the turbine's tilt angle. Additionally, GNSS positioning technology is used to monitor the turbine's position and offset in real time, detecting any deviation or tilt.
[0004] The existing wind turbine safety monitoring systems are generating increasingly large amounts of data. However, these data are only collected, stored, and used for early warning, without sufficient data analysis and mining. They are unable to model and predict the deformation of wind turbines, or provide suggestions and status predictions for the safe operation of wind turbines.
[0005] In conclusion, there is an urgent need for a big data processing method for wind turbine safety monitoring to solve the problems existing in the current technology. Summary of the Invention
[0006] The purpose of this invention is to provide a big data processing method for wind turbine safety monitoring, which can extract the trend characteristics of wind turbine displacement from massive amounts of incomplete and noisy wind turbine monitoring data, analyze the relationship between wind force, wind turbine height and wind turbine displacement, and predict and warn of wind turbine displacement. The specific technical solution is as follows:
[0007] A big data processing method for wind turbine safety monitoring includes:
[0008] Data acquisition: Collect GNSS data, wind data, and IMU data from all monitored wind turbines and store them in the database;
[0009] Real-time calculation of a single wind turbine: GNSS displacement is obtained based on GNSS data, accelerometer displacement is obtained based on IMU data, Kalman filtering is performed on the GNSS displacement and accelerometer displacement to obtain the wind turbine fused displacement; the wind turbine fused displacement is stored in the database and early warning is given based on the wind turbine fused displacement.
[0010] Big data analysis: After establishing the data processing model, the wind turbine height and wind force data of a single wind turbine are input into the data processing model to obtain the displacement model value. Based on the displacement model value, the wind turbine's fused displacement, and the prediction error of the data processing model, abnormal data of the wind turbine are identified. The Kalman gain in the Kalman filter calculation of the wind turbine is dynamically adjusted based on the wind turbine's fused displacement.
[0011] The preferred method for calculating the integrated displacement of the wind turbine in the above technical solutions is as follows:
[0012] (1),
[0013] (2),
[0014] (3),
[0015] in, for t The amount of wind turbine displacement per second. The values represent the difference in displacement and velocity between the GNSS and accelerometer readings. For Kalman gain, For the first t Second observation error, To observe the noise covariance matrix, For accelerometer t The displacement and velocity obtained by the second calculation These are variable parameters that are dynamically adjusted based on the wind turbine's integrated displacement. For prediction t The state vector in seconds, The covariance of displacement and the covariance of velocity output from GNSS and accelerometers. Let be the state transition matrix.
[0016] The preferred system state equation for the Kalman filter among the above technical solutions is:
[0017] (4),
[0018] (5),
[0019] (6),
[0020] in, For the reason t -1 second state prediction t Second state vector, For the first t -1 second actual measured state vector For GNSS in t The displacement and velocity calculated in -1 second For accelerometer t- The displacement and velocity calculated in 1 second For GNSS and accelerometers t The difference in displacement in the eastward direction of -1 second. For GNSS and accelerometers t The difference in northward displacement of -1 second For GNSS and accelerometers t The difference in vertical displacement of -1 second For GNSS and accelerometers t- A 1-second difference in speed in the east direction, For GNSS and accelerometers t The difference in speed in the north direction is -1 second. For GNSS and accelerometers t The difference in vertical velocity of -1 second, 3 3 identity matrices 3 A 3-fold zero matrix.
[0021] The preferred observation equation for the Kalman filter among the above technical solutions is:
[0022] (7),
[0023] (8),
[0024] Among them, the observation vector .
[0025] The preferred approach among the above technical solutions is to establish a data processing model once at interval T1, specifically:
[0026] The wind power data and fused displacement of all wind turbines in the past T3 time period are preprocessed, and the wind turbine height and the preprocessed wind power data and fused displacement are divided into training set and validation set.
[0027] The training of the BP neural network is completed by using the wind power data and wind turbine height in the training set as inputs and the displacement model values as outputs.
[0028] The wind data and turbine height from the validation set are input into the BP neural network, and the prediction error is calculated based on the output displacement model value and the fused displacement of the turbines in the validation set. .
[0029] In the preferred embodiment of the above technical solutions, the abnormal data identification specifically includes:
[0030] Search the database for the combined displacement and wind force data of a single wind turbine within the past T2 time period, and perform preprocessing.
[0031] Input the wind turbine's height and wind force data into the data processing model to obtain the corresponding displacement model value over the past time T2.
[0032] Obtain the difference between the displacement model value corresponding to the past time T2 and the wind turbine fused displacement during the past time T2, and calculate the standard error of the difference. ;
[0033] like If so, it is considered that the data from the past time T2 of the wind turbine is abnormal. n Greater than 1.
[0034] In the preferred embodiment of the above technical solutions, the preprocessing is as follows: first, gross errors are removed from the wind turbine fusion displacement and wind force data using a 3-fold mean square error criterion; then, interpolation algorithms are used to repair missing data; and finally, normalization is performed.
[0035] In the preferred embodiment of the above technical solution, a cluster analysis based on the wind turbine height is performed at an interval of T4. All GNSS data are divided into m categories based on the m wind turbine heights. The fused displacement of the wind turbines in each category is classified into displacement levels. The Kalman gain in the Kalman filter calculation is adjusted according to the displacement level of the wind turbine. Wind turbines whose displacement levels exceed the set threshold are designated as key monitoring wind turbines.
[0036] In the preferred embodiment of the above technical solution, the combined displacement of a single wind turbine is clustered monthly to obtain the combined displacement of the wind turbine in each month, and the month with the largest combined displacement is selected as the key monitoring period. During the key monitoring period, the Kalman gain of the wind turbine, which is confirmed according to the displacement level, is increased by a times, where a > 0.
[0037] In the preferred embodiment of the above technical solution, the acceleration data in the IMU data is integrated into displacement data, and then the displacement data coordinate system is converted to the station center coordinate system using the gyroscope attitude data in the IMU data to obtain the accelerometer displacement within 1 second; the GNSS data of the base station and the monitoring station are used to perform real-time RTK calculation to obtain the displacement of the monitoring station, and then the current displacement obtained by GNSS calculation is subtracted from the displacement of the previous second to obtain the GNSS displacement within 1 second.
[0038] The application of the technical solution of the present invention has the following beneficial effects:
[0039] This invention utilizes GNSS and IMU to monitor wind turbines simultaneously, fuses the data from both, and dynamically adjusts the Kalman gain based on big data analysis results, thereby improving the accuracy and stability of the monitoring results.
[0040] This invention utilizes big data analytics to fully leverage wind turbine monitoring data. It employs cluster analysis to analyze wind turbines of different height categories, extracting those requiring focused attention. The Kalman filter gain parameters in the real-time calculation are dynamically adjusted according to the wind turbine displacement level. After cluster analysis of individual wind turbines, wind turbine displacement and time classifications are obtained, extracting the time periods that require focused monitoring of the wind turbines, and dynamically adjusting the Kalman filter gain parameters.
[0041] This invention utilizes a BP neural network algorithm to obtain the relationship between wind power data, wind turbine height, and the fused displacement of the wind turbine. It uses displacement model values to identify abnormal wind turbines, identify wind turbines with defects, and carry out timely wind turbine maintenance.
[0042] In addition to the objectives, features, and advantages described above, the present invention has other objectives, features, and advantages. The invention will now be described in further detail with reference to the figures. Attached Figure Description
[0043] The accompanying drawings, which form part of this application, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:
[0044] Figure 1 This is a flowchart of the big data processing method for wind turbine safety monitoring;
[0045] Figure 2 This is a flowchart of big data analysis in the big data processing method for wind turbine safety monitoring;
[0046] Figure 3 This is a comparison chart of the displacement model values and the integrated displacement of the wind turbine. Detailed Implementation
[0047] To facilitate understanding of the present invention, a more comprehensive description is provided below, along with preferred embodiments. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the present invention.
[0048] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. Example
[0049] See Figure 1 This embodiment provides a big data processing method for wind turbine safety monitoring, which includes three parts: data acquisition, real-time calculation, and big data analysis. The three parts will be described in detail below.
[0050] Data acquisition: Collect GNSS data, wind data, and IMU data from all monitored wind turbines and store them in the database;
[0051] For specific hardware devices used to acquire data, please refer to existing technologies. Here, IMU refers to an inertial sensor, which includes an accelerometer and a gyroscope. IMU data includes acceleration data acquired by the accelerometer and attitude data acquired by the gyroscope.
[0052] Real-time calculation of a single wind turbine: GNSS displacement is obtained based on GNSS data, accelerometer displacement is obtained based on IMU data, Kalman filtering is performed on the GNSS displacement and accelerometer displacement to obtain the wind turbine fused displacement; the wind turbine fused displacement is stored in the database and early warning is given based on the wind turbine fused displacement.
[0053] Specifically, since the sampling frequency of GNSS data is 1Hz, which is relatively high, while the acquisition frequency of IMU is usually 200Hz, it is necessary to perform preprocessing such as integration and coordinate system transformation on the accelerometer data within 1 second. Specifically, the acceleration data in the IMU data is integrated into displacement data, and then the coordinate system of the displacement data is transformed into the station-centered coordinate system using the gyroscope attitude data in the IMU data to obtain the accelerometer displacement within 1 second.
[0054] Furthermore, the displacement of the monitoring station is obtained by real-time RTK calculation of the GNSS data of the base station and the monitoring station. Then, the displacement of the previous second is subtracted from the current displacement obtained by GNSS calculation to obtain the GNSS displacement within 1 second.
[0055] Preferably, the wind turbine fusion displacement is obtained according to formula (1), and real-time early warning can be achieved by comparing it with the early warning threshold. The calculation method of the wind turbine fusion displacement is as follows:
[0056] (1),
[0057] (2),
[0058] (3),
[0059] in, for t The amount of wind turbine displacement per second. The values represent the difference in displacement and velocity between the GNSS and accelerometer readings. For Kalman gain, For the first t Second observation error, To observe the noise covariance matrix, For accelerometer t The displacement and velocity obtained by the second calculation These are variable parameters that are dynamically adjusted based on the wind turbine's integrated displacement. For prediction t The state vector in seconds, The covariance of displacement and the covariance of velocity output from GNSS and accelerometers. Let be the state transition matrix.
[0060] In this embodiment, the system state equation for the Kalman filter is:
[0061] (4),
[0062] (5),
[0063] (6),
[0064] in, For the reason t -1 second state prediction t Second state vector, For the first t -1 second actual measured state vector For GNSS in t The displacement and velocity calculated in -1 second For accelerometer t- The displacement and velocity calculated in 1 second For GNSS and accelerometers t The difference in displacement in the eastward direction of -1 second. For GNSS and accelerometers t The difference in northward displacement of -1 second For GNSS and accelerometers t The difference in vertical displacement of -1 second For GNSS and accelerometers t- A 1-second difference in speed in the east direction, For GNSS and accelerometers t The difference in speed in the north direction is -1 second. For GNSS and accelerometers t The difference in vertical velocity of -1 second, 3 3 identity matrices 3 A 3-fold zero matrix.
[0065] In this embodiment, the observation equation for the Kalman filter is:
[0066] (7),
[0067] (8),
[0068] Among them, the observation vector .
[0069] Big data analytics: See Figure 2 After establishing the data processing model, the wind turbine height and wind force data of a single wind turbine are input into the data processing model to obtain the displacement model value. Based on the displacement model value, the wind turbine's fused displacement, and the prediction error of the data processing model, abnormal data of the wind turbine are identified. The Kalman gain in the Kalman filter calculation of the wind turbine is dynamically adjusted based on the wind turbine's fused displacement.
[0070] Preferably, the data processing model is established once at interval T1 (in this embodiment, T1 is one month), specifically:
[0071] The wind power data and fused displacement of all wind turbines within the past T3 time period are preprocessed, and the wind turbine height and the preprocessed wind power data and fused displacement of wind turbines are divided into training set and validation set (T3 time is one year in this embodiment).
[0072] The training of the BP neural network is completed by using the wind power data and wind turbine height in the training set as inputs and the displacement model values as outputs.
[0073] The wind data and turbine height from the validation set are input into the BP neural network, and the prediction error is calculated based on the output displacement model value and the fused displacement of the turbines in the validation set. .
[0074] Specifically, the prediction error The calculation method is as follows:
[0075] (9),
[0076] In equation (9), For displacement model values, This represents the combined displacement of the wind turbine.
[0077] Specifically, in this embodiment, the input layer nodes of the BP neural network are wind power data and wind turbine height, the number of hidden layer nodes is 8, and the output layer is displacement model values.
[0078] After the data processing model is established, abnormal data identification can be performed, specifically:
[0079] The database is searched for the combined displacement and wind force data of a single wind turbine over the past T2 time period, and preprocessed (T2 time period is 2 months in this embodiment).
[0080] Input the wind turbine's height and wind force data into the data processing model to obtain the corresponding displacement model value over the past time T2.
[0081] Obtain the difference between the displacement model value corresponding to the past time T2 and the wind turbine fused displacement during the past time T2, and calculate the standard error of the difference. ;
[0082] like If this is the case, then the data from the past time T2 of the wind turbine is considered abnormal, and this abnormality is output in the report. Preferably, the... n >1, in this embodiment n =3.
[0083] Furthermore, in practical applications, the values of T1, T2, and T3 are not limited to those provided in this embodiment. Those skilled in the art can modify them according to the actual monitoring situation. In order to ensure the accuracy of the data processing model, T3 > T2 > T1 in general.
[0084] like Figure 3 In the wind turbine result diagram shown, the circle in the middle represents the model result, and the ellipses at both ends represent the abnormal results. It can be seen that there is too much abnormal data for this wind turbine, which requires special attention.
[0085] Preferably, the preprocessing in this embodiment involves: firstly, using a 3x mean square error criterion to remove gross errors from the fused displacement and wind force data of the wind turbine; then, using interpolation algorithms to repair missing data; and finally, performing normalization. Preprocessing reduces noise interference and data loss, ensuring the accuracy of comparisons and analyses between different parameters. The interpolation algorithm can employ polynomial fitting or inverse distance weighting.
[0086] Furthermore, in this embodiment, the analysis results are visualized through cluster analysis based on different wind turbine height categories and cluster analysis for individual wind turbines.
[0087] The clustering analysis for different wind turbine height categories is as follows: A clustering analysis based on wind turbine height is performed at intervals T4. All GNSS data are divided into m categories based on the m wind turbine heights. For each category, the fused displacement of the wind turbines is classified into displacement levels using the k-medoids algorithm to achieve risk assessment. The Kalman gain in the Kalman filter calculation is adjusted according to the displacement level corresponding to the wind turbine; specifically, the parameters are adjusted. Meanwhile, fans whose displacement levels exceed the set threshold will be designated as key fans for monitoring.
[0088] In this embodiment, based on different displacement levels, refer to Table 1 for... Parameters are dynamically adjusted:
[0089] Table 1
[0090]
[0091] Preferred, The initial values and the wind turbines without displacement classification, All values are set to 1. In this embodiment, T4 is set to one month, but the value of T4 can also be adjusted according to the actual monitoring situation.
[0092] Cluster analysis for individual wind turbines involves performing monthly cluster analysis on the combined displacement of each turbine. The month with the largest combined displacement is selected as the key monitoring period. Within this key monitoring period, the Kalman gain, determined based on the displacement level, is increased by a factor of *a*, i.e., the parameter... After adding 'a', we resubmit it into formula (3), where 'a' > 0. In this embodiment, 'a' is 0.2.
[0093] After the big data analysis is completed, reports are output to visualize the analysis results, including the key monitoring months for each wind turbine, the key monitoring turbine number, and the wind turbine model, providing an intuitive basis for adjusting and maintaining wind turbine safety monitoring strategies.
[0094] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A big data processing method for wind turbine safety monitoring, characterized in that, include: Data acquisition: Collect GNSS data, wind data, and IMU data from all monitored wind turbines and store them in the database; Real-time calculation of a single wind turbine: GNSS displacement is obtained based on GNSS data, accelerometer displacement is obtained based on IMU data, Kalman filtering is performed on the GNSS displacement and accelerometer displacement to obtain the wind turbine fused displacement; the wind turbine fused displacement is stored in the database and early warning is given based on the wind turbine fused displacement. Big data analysis: After establishing the data processing model, the wind turbine height and wind force data of a single wind turbine are input into the data processing model to obtain the displacement model value. Based on the displacement model value, the wind turbine's fused displacement, and the prediction error of the data processing model, abnormal data of the wind turbine are identified. The Kalman gain in the Kalman filter calculation of the wind turbine is dynamically adjusted based on the wind turbine's fused displacement.
2. The big data processing method for wind turbine safety monitoring according to claim 1, characterized in that, The calculation method for the integrated displacement of the wind turbine is as follows: （1）， （2）， （3）， in, for t The amount of wind turbine displacement per second. The values represent the difference in displacement and velocity between the GNSS and accelerometer readings. For Kalman gain, For the first t Second observation error, To observe the noise covariance matrix, For accelerometer t The displacement and velocity obtained by the second calculation These are variable parameters that are dynamically adjusted based on the wind turbine's integrated displacement. For prediction t The state vector in seconds, The covariance of displacement and the covariance of velocity output from GNSS and accelerometers. Let be the state transition matrix.
3. The big data processing method for wind turbine safety monitoring according to claim 2, characterized in that, The system state equation for Kalman filtering is: （4）， （5）， （6）， in, For the reason t -1 second state prediction t Second state vector, For the first t -1 second actual measured state vector For GNSS in t The displacement and velocity obtained by solving in -1 second For accelerometer t- The displacement and velocity calculated in 1 second For GNSS and accelerometers t The difference in displacement in the eastward direction of -1 second. For GNSS and accelerometers t The difference in northward displacement of -1 second For GNSS and accelerometers t The difference in vertical displacement of -1 second For GNSS and accelerometers t- A 1-second difference in speed in the east direction, For GNSS and accelerometers t The difference in speed in the north direction is -1 second. For GNSS and accelerometers t The difference in vertical velocity of -1 second, 3 3 identity matrices 3 A 3-fold zero matrix.
4. The big data processing method for wind turbine safety monitoring according to claim 3, characterized in that, The observation equation for Kalman filtering is: （7）， （8）， Among them, the observation vector .
5. The big data processing method for wind turbine safety monitoring according to claim 1, characterized in that, The data processing model is established once at interval T1, specifically as follows: The wind power data and fused displacement of all wind turbines in the past T3 time period are preprocessed, and the wind turbine height and the preprocessed wind power data and fused displacement are divided into training set and validation set. The training of the BP neural network is completed by using the wind power data and wind turbine height in the training set as inputs and the displacement model values as outputs. The wind data and turbine height from the validation set are input into the BP neural network, and the prediction error is calculated based on the output displacement model value and the fused displacement of the turbines in the validation set. .
6. The big data processing method for wind turbine safety monitoring according to claim 5, characterized in that, The specific steps of abnormal data identification are: Search the database for the combined displacement and wind force data of a single wind turbine within the past T2 time period, and perform preprocessing. Input the wind turbine's height and wind force data into the data processing model to obtain the corresponding displacement model value over the past time T2. Obtain the difference between the displacement model value corresponding to the past time T2 and the wind turbine fused displacement value of the past time T2, and calculate the standard error of the difference. ; like If so, it is considered that the data from the past time T2 of the wind turbine is abnormal. n Greater than 1.
7. The big data processing method for wind turbine safety monitoring according to claim 6, characterized in that, The preprocessing is as follows: first, gross errors are removed from the wind turbine fusion displacement and wind force data using the 3x mean square error criterion; then, interpolation algorithms are used to repair missing data; and finally, normalization is performed.
8. The big data processing method for wind turbine safety monitoring according to claim 1, characterized in that, Cluster analysis based on wind turbine height is performed at interval T4. All GNSS data are divided into m classes based on the height of m wind turbines. The fused displacement of wind turbines in each class is classified into displacement levels. The Kalman gain in the Kalman filter calculation is adjusted according to the displacement level of the wind turbine. Wind turbines with displacement levels exceeding the set threshold are designated as key monitoring wind turbines.
9. The big data processing method for wind turbine safety monitoring according to claim 7, characterized in that, Cluster analysis was performed on the combined displacement of a single wind turbine on a monthly basis to obtain the combined displacement of the wind turbine in each month. The month with the largest combined displacement was selected as the key monitoring period. During the key monitoring period, the Kalman gain of the wind turbine, which was determined based on the displacement level, was increased by a times, where a > 0.
10. The big data processing method for wind turbine safety monitoring according to claim 1, characterized in that: The acceleration data in the IMU data is integrated into displacement data, and then the displacement data coordinate system is converted to the station center coordinate system using the gyroscope attitude data in the IMU data to obtain the accelerometer displacement within 1 second. The displacement of the monitoring station is obtained by real-time RTK calculation of the GNSS data of the base station and the monitoring station. Then, the displacement of the previous second is subtracted from the current displacement obtained by GNSS calculation to obtain the GNSS displacement within 1 second.