Wind power tower structure state monitoring method and system

Abstract

The invention discloses a wind power tower structure state monitoring method and system, and belongs to the technical field of wind power tower monitoring, and the method comprises the steps: obtaining historical monitoring data of the top of a wind power tower, and carrying out the preprocessing of the data, and obtaining a weight set and a verification set; performing principal component analysis on the data in the weight set to obtain a data contribution degree; constructing an anomaly evaluation model based on the data contribution degree, and verifying the anomaly evaluation model by using the verification set to obtain an accurate evaluation model; acquiring real-time monitoring data of the top of the wind power tower, and inputting the real-time monitoring data into the accurate evaluation model to obtain an evaluation result; a corresponding wind power tower structure state monitoring system responds to the evaluation result to execute an alarm action; according to the method, the generality of the mass monitoring data is used as the basis of the accurate evaluation model through the data contribution degree, then the accurate evaluation model is used for rapidly processing the mass monitoring data, and the real-time performance of wind power tower monitoring is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wind turbine tower monitoring, and specifically to a method and system for monitoring the structural state of a wind turbine tower. Background Art

[0002] As the service life of wind power generation devices increases, corrosion, rusting and other events occur in the corresponding wind turbine towers. After such events occur, they will cause damage to the flange plates of the wind turbine towers, loosening of bolts, and breakage of welds, etc., making the daily operation of the wind turbine towers have potential safety hazards and even collapse; in order to detect the potential safety hazards in advance and prevent the occurrence of collapse accidents, the prior art generally uses a lot of sensors to monitor the key parts of the wind turbine tower, but this requires spending a lot of time to process a large amount of monitoring data, or uses methods such as video monitoring and image recognition, but this is extremely vulnerable to factors such as weather, resulting in recognition failure, and the hidden danger can only be discovered after the wind turbine tower is significantly tilted.

[0003] Chinese Patent, Publication No.: CN113107786A, Publication Date: July 13, 2021, discloses a method, device and equipment for monitoring the safety of a wind turbine tower flange plate, including: obtaining in real time the strain data sensed by each dynamic strain gauge arranged on the flange plate and the vibration speed sensed by each vibration sensor arranged on the flange plate; calculating the gap displacement corresponding to each dynamic strain gauge according to the strain data sensed by each dynamic strain gauge and the sensitive length of each dynamic strain gauge; analyzing the gap displacements corresponding to each dynamic strain gauge to determine the target gap safety factor of the flange plate; analyzing the vibration speeds sensed by each vibration sensor to determine the target vibration safety factor of the flange plate; evaluating the safety performance of the flange plate according to the target gap safety factor, the target vibration safety factor and a pre-set safety performance evaluation rule to obtain the safety performance evaluation information of the flange plate; however, this invention does not consider the commonality of a large amount of monitoring data, resulting in low processing efficiency of the monitoring data, and further resulting in poor real-time performance of the corresponding wind turbine tower monitoring. Summary of the Invention

[0004] The object of the present invention is to address the problem in the prior art that the commonality of a large amount of monitoring data is not considered, resulting in poor real-time performance of the corresponding wind turbine tower monitoring. A wind turbine tower structural state monitoring method is proposed. The historical monitoring data at the top of the wind turbine tower is preprocessed to obtain a weight set and a validation set. Principal component analysis is performed on the data in the weight set to obtain the data contribution degree. An anomaly evaluation model is constructed based on the data contribution degree, and the anomaly evaluation model is verified using the validation set to obtain an accurate evaluation model. The real-time monitoring data of the wind turbine tower is input into the accurate evaluation model to obtain an evaluation result. Finally, the corresponding wind turbine tower structural state monitoring system performs an alarm action in response to the evaluation result. By using the data contribution degree, the commonality of the large amount of monitoring data is used as the basis for the accurate evaluation model, and thus the accurate evaluation model can be used to quickly process the large amount of monitoring data, effectively improving the real-time performance of the wind turbine tower monitoring.

[0005] In a first aspect, a technical solution provided in an embodiment of the present invention is a wind turbine tower structural state monitoring method, including the following steps: Obtain the historical monitoring data at the top of the wind turbine tower and perform preprocessing to obtain a weight set and a validation set; Perform principal component analysis on the data in the weight set to obtain the data contribution degree; Construct an anomaly evaluation model based on the data contribution degree, and verify the anomaly evaluation model using the validation set to obtain an accurate evaluation model; obtain the real-time monitoring data at the top of the wind turbine tower, and input the real-time monitoring data into the accurate evaluation model to obtain an evaluation result; The corresponding wind turbine tower structural state monitoring system performs an alarm action in response to the evaluation result.

[0006] In this solution, historical monitoring data of the wind turbine tower top including at least the tilt angle, vibration displacement, and vibration acceleration are obtained. The tilt angle, vibration displacement, and vibration acceleration are all related to the vibration intensity of the wind turbine tower top, and the vibration intensity is related to flange damage, bolt loosening, weld breakage, etc. Then, it is possible to determine whether there are potential safety hazards in the corresponding wind turbine tower through data such as the tilt angle, vibration displacement, and vibration acceleration. Preprocessing the historical monitoring data of the wind turbine tower top to obtain a weight set and a validation set can eliminate abnormal data and missing data caused by force majeure factors such as external factors, and can also unify different types of data into a reasonable range for easy processing. Performing principal component analysis on the data in the weight set to obtain the data contribution degree. Combining the fact that the tilt angle, vibration displacement, and vibration acceleration are all related to the vibration intensity of the wind turbine tower top, the data contribution degree is essentially the contribution degree of the tilt angle, vibration displacement, and vibration acceleration to the vibration intensity respectively, that is, the correlation degree of different types of data in the weight set to the potential safety hazards of the wind turbine tower. Furthermore, an abnormal evaluation model can be established based on the correlation degree, so that the abnormal evaluation model can directly use data such as the tilt angle, vibration displacement, and vibration acceleration to evaluate the vibration intensity of the corresponding wind turbine tower, realizing the monitoring of the potential safety hazards of the wind turbine tower. However, the performance of the abnormal evaluation model may be relatively poor. Using the validation set to verify the abnormal evaluation model to obtain an accurate evaluation model, and adjusting the constraint conditions of the abnormal evaluation model by taking the validation set as a benchmark, ensuring the performance of the abnormal evaluation model. Inputting the real-time monitoring data into the accurate evaluation model to obtain an evaluation result. The evaluation result at least includes the monitoring data and the status of the monitoring data, that is, if the wind turbine tower corresponding to the monitoring data has potential safety hazards, the monitoring data is abnormal monitoring data, and if the wind turbine tower corresponding to the monitoring data has no potential safety hazards, the monitoring data is normal monitoring data. Finally, the corresponding wind turbine tower structure state monitoring system performs an alarm action in response to the evaluation result. Using the commonality of the massive monitoring data as the basis of the accurate evaluation model through the data contribution degree, and then the accurate evaluation model can be used to quickly process the massive monitoring data, effectively improving the real-time performance of wind turbine tower monitoring.

[0007] Preferably, the specific process of preprocessing the historical monitoring data of the wind turbine tower top to obtain a weight set and a validation set is as follows: Statistically analyze the outliers and missing values in the historical monitoring data of the wind turbine tower top, and delete the outliers to obtain a missing data set; Fill in the missing data in the missing data set based on the interpolation method to obtain a complete data set, and normalize the data in the complete data set based on the normalization criterion to obtain a standard data set; Divide the standard data set based on a preset weight-validation ratio to obtain a weight set and a validation set.

[0008] Preferably, the specific process of performing principal component analysis on the data in the weight set to obtain the data contribution degree is as follows: Decentralize the data based on the mean of the data in the weight dataset to obtain decentralized data; Construct a feature matrix based on the decentralized data, and transform the feature matrix into a covariance matrix based on the covariance matrix transformation criterion; Decompose the covariance matrix based on the eigenvalue solution criterion to obtain the principal component direction and variance contribution rate; Determine the principal components based on the variance contribution rate and the principal component direction, and calculate the contribution degree of the data in the weight dataset based on the principal components to obtain the data contribution degree.

[0009] In this solution, the essence of decentralization is to subtract the mean of different types of data from each type of data. If there are a total of 5 groups of data, and each group of data contains three types of data: tilt angle, vibration displacement, and vibration acceleration. Taking the tilt angle as an example, the tilt angles in the 5 groups of data need to be extracted, and the average tilt angle is calculated. Then, each tilt angle is subtracted by the average tilt angle, which is the entire process of decentralization. The completely decentralized data is filled into the original data positions to obtain the decentralized data. Taking the first group of data as the first row and the fifth group of data as the fifth row, the 5 groups of data are combined in sequence to obtain a feature matrix. In order to accurately analyze the data characteristics of the monitoring data, the feature matrix needs to be transformed into a covariance matrix based on the covariance matrix transformation criterion. Then, after solving the covariance matrix, the corresponding eigenvalues and eigenvectors can be obtained. Based on the fact that the eigenvalues, eigenvectors, and covariance are directly related, and the tilt angle, vibration displacement, and vibration acceleration are all related to the vibration intensity of the wind turbine tower, it can be inferred that the eigenvectors contain the principal components that affect the vibration intensity. Therefore, the eigenvectors are used as the principal component directions. The eigenvalues themselves are related to the eigenvectors, and the magnitude of the eigenvalue data is associated with the eigenvectors. Then, the eigenvalues are related to the variance contribution rate. Considering that the corresponding monitoring data has been standardized in the preprocessing process, if there are only three types of data, namely tilt angle, vibration displacement, and vibration acceleration, in the monitoring data, the total variance contribution rate, that is, the total variance contribution rate is equal to the number of types of monitoring data, and the total variance contribution rate is 3. The variance contribution rate can be obtained by dividing the corresponding eigenvalue by the total variance contribution rate. Then, the principal components can be determined in the principal component direction using the variance contribution rate, and the data contribution degrees of different types of data can be calculated based on the contribution degree formula and the principal components.

[0010] Preferably, the mathematical model of the covariance matrix transformation criterion is specifically as follows: In the formula, C is the covariance matrix, X is the feature matrix, and X T is the transpose matrix of the feature matrix, and n is the number of samples corresponding to the decentralized data.

[0011] Preferably, the specific process of determining the principal components based on the variance contribution rate and the principal component direction is as follows: Sort the variance contribution rates of all undetermined principal components in the principal component direction in descending order, and accumulate the variance contribution rates starting from the largest one to obtain the cumulative variance contribution rate; When the cumulative variance contribution rate is greater than or equal to the preset contribution rate threshold, stop accumulating the variance contribution rates, and mark the undetermined principal component corresponding to the cumulative variance contribution rate as the principal component.

[0012] In this solution, since the variance contribution rate is related to the covariance matrix and the eigenvalues are different in size, the variance contribution rates are also different in size. Therefore, the variance contribution rates can be sorted in descending order. The larger the value of the variance contribution rate, the greater the influence of the corresponding type of data on the whole. That is, the variance contribution rates of the tilt angle, vibration displacement, and vibration acceleration respectively represent their own influences on the vibration intensity of the wind turbine tower. Therefore, the cumulative variance contribution rate can be obtained by accumulating the variance contribution rates starting from the largest one in turn, and the principal components can be screened by using the contribution rate threshold and the cumulative variance contribution rate. When the cumulative variance contribution rate is greater than or equal to the contribution rate threshold, it proves that the corresponding principal component can already affect the vibration intensity of the wind turbine tower on the premise of ignoring the remaining other principal components, and the remaining other principal components cannot weaken this influence.

[0013] Preferably, the specific contribution degree formula corresponding to the data contribution degree is as follows: In the formula, Y i is the data contribution degree of the i-th centered data, α ij is the loading of the i-th centered data on the j-th principal component, y j is the variance contribution rate of the j-th principal component, and p is the total number of principal components.

[0014] In this solution, the loading is determined by the loading matrix, and the loading matrix is essentially the eigenvector matrix obtained by solving the covariance matrix.

[0015] Preferably, the specific process of constructing an anomaly evaluation model based on the data contribution degree and using the validation set to verify the anomaly evaluation model to obtain an accurate evaluation model is as follows: Set the weights of the data in the corresponding weight dataset based on the data contribution degree, and extract the data features of the data in the weight dataset; Construct an anomaly evaluation model based on the weights and data features, and input the weight set into the anomaly evaluation model to obtain a weight result; Compare and analyze the weight result with the validation set to obtain a validation difference, and adjust the anomaly evaluation model based on the validation difference; Mark the anomaly evaluation model that has completed adjustment as the accurate evaluation model.

[0016] In this solution, the data contribution degree is converted into a number between 0 and 1, and it is required that the sum of the converted data contribution degrees is equal to 1. If the sum of the converted data contribution degrees is not equal to 1, a constant term weight is added for adjustment, and the mean and standard deviation corresponding to the data in the weight set are statistically obtained, that is, the data characteristics. Based on the weights and the types of monitoring data, an objective function is established. The mean minus a certain multiple of the standard deviation is used as the judgment threshold, and the multiple is set according to experience. The judgment interval established based on the judgment threshold is used as a constraint condition. The objective function and the constraint condition are sorted out to obtain an anomaly evaluation model. The weight set is input into the anomaly evaluation model to obtain a weight result. The weight result reflects the performance of the anomaly evaluation model in processing data. By comparing and analyzing the weight result with the validation set, a validation difference can be obtained. The performance of the anomaly evaluation model can be quantified using the validation difference, and the accuracy of the anomaly evaluation model in using monitoring data to judge whether there are potential safety hazards in the structural state of the wind turbine tower can be obtained. Furthermore, the threshold in the constraint condition can be adjusted to effectively improve the accuracy of the anomaly evaluation model.

[0017] Preferably, the specific process of comparing and analyzing the weight result with the validation set to obtain the validation difference is as follows: Align the validation set based on the time series corresponding to the weight result to obtain a common time series, and input the aligned validation set into the anomaly evaluation model to obtain a validation result; Plot the data corresponding to the weight result as a weight curve graph, and plot the data corresponding to the validation result as a validation curve graph; Integrate the weight curve graph and the validation curve graph to obtain a comparison curve graph; Based on the common time series, divide the curve interval of the comparison curve graph, and extract the weight trend and validation trend in the curve interval; Compare the weight trend with the validation trend, and sort out the time periods, weight sets, and validation sets where the weight trend direction is different from the validation trend direction to obtain the validation difference.

[0018] In this solution, the time series of the weight set is aligned with the time series of the validation set, so that the weight results corresponding to the weight set and the validation results corresponding to the validation set are highly consistent in time. Furthermore, the weight curve and the validation curve are highly consistent in time. The weight curve and the validation curve are plotted in the same graph to obtain a comparison curve graph. At this time, the time series of the weight curve is the same as the time series of the validation curve. Then, the time series of the comparison curve graph is essentially a common time series. Since the data in the weight results and the validation results correspond to two types of data, normal data and abnormal data, in order to comprehensively analyze the differences between the data and the corresponding data at the same time in history, the normal data is divided into one or more intervals based on the common time series, and the abnormal data is divided into one or more intervals. The weight trend and the validation trend in the curve intervals are extracted, and the weight trend is compared with the validation trend to obtain the differences between the data and the corresponding data at the same time in history, that is, the time periods when the weight trend direction is the same as the validation trend direction and the time periods when the weight trend direction is different from the validation trend direction. Then, the time periods when the weight trend direction is different from the validation trend direction, the weight set, and the validation set are sorted out to obtain the validation differences.

[0019] Preferably, the specific process for the corresponding wind turbine tower structure state monitoring system to perform an alarm action in response to the evaluation result is as follows: Extract the real-time monitoring data determined to be abnormal in the evaluation result, and determine the abnormal wind turbine tower based on the real-time monitoring data; extract the operation and maintenance information and operation and maintenance phone numbers of the corresponding abnormal wind turbine tower from the operation and maintenance information database, and convert the operation and maintenance information into a short message and send it to the operation and maintenance phone number.

[0020] On the other hand, a technical solution provided in the embodiments of the present invention is a wind turbine tower structure state monitoring system, including: a data acquisition module, a principal component module, an evaluation module, and an alarm module; The data acquisition module is used to collect and preprocess the monitoring data at the top of the wind turbine tower to obtain a weight set, a validation set, and real-time monitoring data, transmit the weight set to the principal component module, and transmit the real-time monitoring data and the validation set to the evaluation module; The principal component module performs principal component analysis on the weight set uploaded by the data acquisition module to obtain the data contribution degree, and transmits the data contribution degree to the evaluation module; The evaluation module constructs an abnormal evaluation model based on the data contribution degree uploaded by the principal component module, validates the abnormal evaluation model based on the validation set uploaded by the data acquisition module to obtain an accurate evaluation model, and transmits the evaluation result obtained by inputting the real-time monitoring data uploaded by the data acquisition module into the accurate evaluation model to the alarm module; The alarm module performs an alarm action based on the evaluation result uploaded by the evaluation module.

[0021] The beneficial effects of the present invention: (1) The present application performs principal component analysis on the historical monitoring data of a wind turbine tower to obtain the data contribution degree. Considering that the tilt angle, vibration displacement, and vibration acceleration in the historical monitoring data are all related to the vibration intensity at the top of the wind turbine tower, and the vibration intensity can intuitively reflect the overall structural stability of the wind turbine tower, the data contribution degree can represent the correlation degree between different types of data in the monitoring data and the vibration intensity. Furthermore, an anomaly evaluation model can be established based on this correlation degree, enabling the anomaly evaluation model to directly use data such as tilt angle, vibration displacement, and vibration acceleration to evaluate the vibration intensity of the corresponding wind turbine tower, thereby realizing the monitoring of potential safety hazards of the wind turbine tower. (2) The present application uses a validation set to verify the anomaly evaluation model to obtain an accurate evaluation model. By adjusting the constraint conditions of the anomaly evaluation model with the validation set as a benchmark, the performance of the anomaly evaluation model is ensured. Then, the real-time monitoring data is input into the accurate evaluation model to obtain an evaluation result, which at least includes the monitoring data and the status of the monitoring data. That is, if there are potential safety hazards in the wind turbine tower corresponding to the monitoring data, the monitoring data is abnormal monitoring data; if there are no potential safety hazards in the wind turbine tower corresponding to the monitoring data, the monitoring data is normal monitoring data. Finally, the corresponding wind turbine tower structure state monitoring system performs an alarm action in response to the evaluation result. By using the data contribution degree, the commonality of the massive monitoring data is used as the basis of the accurate evaluation model. Furthermore, the accurate evaluation model can be used to quickly process the massive monitoring data, effectively improving the real-time performance of wind turbine tower monitoring. Description of the Drawings

[0022] By reading the detailed description of the non-limiting embodiments with reference to the following drawings, other features, objectives, and advantages of the present invention will become more apparent. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention. Moreover, throughout the drawings, the same reference numerals are used to represent the same components.

[0023] Figure 1 It is a schematic flowchart of a method for monitoring the structure state of a wind turbine tower; Figure 2 It is a schematic structural diagram of a system for monitoring the structure state of a wind turbine tower. Detailed Embodiments

[0024] To make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only the best embodiments of the present invention and are only used to explain the present invention, without limiting the protection scope of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0025] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts depict the operations (or steps) as sequential processes, many of the operations (or steps) can be implemented in parallel, concurrently, or simultaneously. In addition, the order of the operations can be rearranged. The process can be terminated when its operations are completed, but it can also have additional steps not included in the figures; the process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0026] Embodiment 1: As Figure 1 shown, this embodiment provides a method for monitoring the structural state of a wind turbine tower, including the following steps: Obtain the historical monitoring data of the top of the wind turbine tower and perform preprocessing to obtain a weight set and a validation set; Specifically, count the outliers and missing values in the historical monitoring data of the top of the wind turbine tower, and delete the outliers to obtain a missing data set; Fill the missing data in the missing data set based on the interpolation method to obtain a complete data set, and normalize the data in the complete data set based on the normalization criterion to obtain a standard data set; Divide the standard data set based on a preset weight-validation ratio to obtain a weight set and a validation set.

[0027] In this embodiment, according to the sway law of the wind turbine tower barrel and the vibration law of the wind turbine tower, a yaw platform or a saddle platform near the nacelle at the top of the tower is selected to install sensors. The sensors are used to collect the monitoring data of the top of the wind turbine tower. The monitoring data includes at least the tilt angle, vibration displacement, and vibration acceleration. To monitor the movement trajectory of the top of the wind turbine tower in detail, one monitoring point, three monitoring parameters, and six monitoring directions are set. The monitoring point is used to install sensors. The monitoring parameters are essentially the tilt angle, vibration displacement, and vibration acceleration. The monitoring directions are set with the main wind direction of the wind farm as the reference direction. After collecting the data, count the outliers and missing values in the data, and delete the outliers so that only missing values exist in the overall data. Considering that the vibration of the top of the wind turbine tower is regular, the monitoring data correspondingly has a linear relationship. The missing data can be filled based on the interpolation method combined with the linear relationship to obtain a complete data set, that is, a complete data set. To eliminate the influence of different data dimensions on the subsequent data processing process, the data in the complete data set is normalized based on the normalization criterion to obtain a standard data set. The mathematical expression of the normalization criterion is: In the formula, X b is the data after normalization, x is the data before normalization, μ is the mean of the data, and σ is the standard deviation of the data; Next, a weight set and a validation set are obtained based on a preset weight-validation ratio for dividing the standard data set. The weight set accounts for 90% of the standard data set, and the validation set accounts for 10% of the standard data set.

[0028] Perform principal component analysis on the data in the weight set to obtain the data contribution degree; Specifically, based on the mean of the data in the weight data set, the data is decentralized to obtain decentralized data; Construct a feature matrix based on the decentralized data, and convert the feature matrix into a covariance matrix based on the covariance matrix conversion criterion; the mathematical model of the covariance matrix conversion criterion is specifically: In the formula, C is the covariance matrix, X is the feature matrix, X T is the transposed matrix of the feature matrix, and n is the number of samples corresponding to the decentralized data; Decompose the covariance matrix based on the eigenvalue solving criterion to obtain the principal component direction and the variance contribution rate; Determine the principal components based on the variance contribution rate and the principal component direction; Sort the variance contribution rates of all the to-be-determined principal components in the principal component direction in descending order, and start accumulating from the largest variance contribution rate to obtain the cumulative variance contribution rate; When the cumulative variance contribution rate is greater than or equal to the preset contribution rate threshold, stop the accumulation of the variance contribution rate, and mark the to-be-determined principal components corresponding to the cumulative variance contribution rate as the principal components; Calculate the contribution degree of the data in the weight data set based on the principal components to obtain the data contribution degree; The contribution degree formula corresponding to the data contribution degree is specifically: In the formula, Y i is the data contribution degree of the i-th type of decentralized data, α ij is the load of the i-th type of decentralized data on the j-th principal component, y j is the variance contribution rate of the j-th principal component, and p is the total number of principal components.

[0029] In this embodiment, the weight set includes at least the tilt angle, the vibration displacement, and the vibration acceleration. Then, the data in the weight set can be represented by Table 1, and Table 1 is specifically: Table 1. Data Table of the Weight Set sample tilt angle vibration displacement vibration acceleration 1 0.30° 0.05 mm <![CDATA[2.12m / s 2 > 2 0.64° 0.12 mm <![CDATA[2.96m / s 2 > 3 0.85° 0.17 mm <![CDATA[3.77m / s 2 > Among them, the sample is actually the number of acquisitions of the sensor. Each time the sensor acquires data, it will correspondingly acquire the tilt angle, the vibration displacement, and the vibration acceleration; in order to decentralize the data in Table 1, it is necessary to calculate the mean of different types of data based on the mean formula, and the mean formula is specifically: In the formula, is the mean of data m, where m i is the i-th data m, m is the type of data, and M is the number of data m; By calculation, the mean of the tilt angle is 0.597°, the mean of the vibration displacement is 0.113 mm, and the mean of the vibration acceleration is 2.95 m / s 2 , then perform a centering operation on the data, subtract the corresponding mean from the data in each sample to obtain Table 2, and the specific content of Table 2 is: Table 2. Centered Data Then, the feature matrix x constructed based on the centered data centered can be specifically expressed as: The covariance matrix C obtained by converting the feature matrix based on the covariance matrix conversion criterion is specifically: The covariance matrix is a 3-order matrix. Use the matrix solution rules to solve the covariance matrix to obtain the eigenvalues λ 1 = 0.76, and the eigenvector c 1 = (0.72 -0.02 -0.70) T , the eigenvalue λ 2 = 0.0004, and the eigenvector c 2 = (0.69-0.08 -0.70) T , the eigenvalue λ 3 = 0.0004, and the eigenvector c 3 = (0.69 -0.08 -0.70) T . Take the eigenvector as the principal component direction, then there are corresponding principal component directions c 1 , principal component direction c 2 , principal component direction c 3 ; Secondly, there are a total of 3 eigenvalues. The eigenvalues can be input into the variance contribution rate formula to obtain the variance contribution rate. The specific variance contribution rate formula is: In the formula, λ f,i represents the variance contribution rate corresponding to the i-th eigenvalue, and λ i is the i-th eigenvalue; Then there are a total of 3 variance contribution rates, which are λ f,1 = 0.999, λ f,2 = 0.0005, λf,3 = 0.0005. Based on the variance contribution rate, the eigenvalue λ f,1 The corresponding tilt angle is sufficient to determine whether there are potential safety hazards in the working environment of the corresponding wind turbine tower. The working environment may be a strong wind environment that has lasted for some time, and the collected monitoring data does not include the data of large vibration displacement of the wind turbine tower at the beginning of the strong wind. Then, the variance contribution rate λ that will obviously exceed the contribution rate threshold f,1 The corresponding tilt angle is marked as the principal component, and the corresponding contribution degree Y is calculated based on the contribution degree formula 1 = 0.518, that is, 51.8%.

[0030] Construct an anomaly evaluation model based on the data contribution degree, and use the validation set to verify the anomaly evaluation model to obtain an accurate evaluation model; Set the weights of the data in the corresponding weight dataset based on the data contribution degree, and extract the data features of the data in the weight set; Construct an anomaly evaluation model based on the weights and data features, and input the weight set into the anomaly evaluation model to obtain a weight result; compare the weight result with the validation set to obtain a validation difference; Align the time series of the weight set corresponding to the weight result with the validation set to obtain a common time series, and input the aligned validation set into the anomaly evaluation model to obtain a validation result; Plot the data corresponding to the weight result as a weight curve graph, and plot the data corresponding to the validation result as a validation curve graph; Integrate the weight curve graph and the validation curve graph to obtain a comparison curve graph; Based on the common time series, divide the curve interval of the comparison curve graph, and extract the weight trend and validation trend in the curve interval; compare the weight trend with the validation trend, and sort out the time periods, weight sets, and validation sets where the weight trend and validation trend are different to obtain a validation difference; And adjust the anomaly evaluation model based on the validation difference; Mark the adjusted anomaly evaluation model as an accurate evaluation model.

[0031] In this embodiment, the data contribution degree is converted into a number between 0 and 1, and it is required that the sum of the converted data contribution degrees is equal to 1. If the sum of the converted data contribution degrees is not equal to 1, a constant term weight is added for adjustment. Since it has been determined in the process of principal component analysis that in the current working environment, only the tilt angle can be used to detect the potential safety hazards of the corresponding wind turbine tower, that is, the structural state of the wind turbine tower, the data contribution degree Y 1 = 51.8% is converted into 0.518, and a constant term weight of 0.482 is added. Based on the weights, an objective function is established in combination with the types of monitoring data. The specific objective function is: f = 0.518Y1,f +0.482Y 0,f ; where f is the vibration intensity at the top of the corresponding wind turbine tower, and Y 1,f is the tilt angle in the monitoring data at the top of the wind turbine tower, and Y 0,f is a constant, and the constant needs to be set according to the historical monitoring data at the top of the wind turbine tower to simulate the interference during the monitoring process at the top of the wind turbine tower.

[0032] Next, the data in the weight set is statistically analyzed to obtain the corresponding mean value and standard deviation, that is, the data characteristics. The mean value minus a certain multiple of the standard deviation is used as the judgment threshold α, and the multiple is set according to experience. The judgment interval established based on the judgment threshold is used as a constraint condition, and the constraint condition can be expressed as -α ≤ f ≤ α. The objective function and the constraint condition are sorted out to obtain an anomaly evaluation model. The weight set is input into the anomaly evaluation model to obtain a weight result. Combining with the objective function, it is actually to input the tilt angle in the weight set into the anomaly evaluation model to obtain the vibration intensity of the corresponding wind turbine tower. The weight result reflects the performance of the anomaly evaluation model in processing data. The verification difference is obtained by comparing and analyzing the weight result with the verification set.

[0033] Specifically, align the time series of the weight set with the time series of the verification set so that the weight result corresponding to the weight set and the verification result corresponding to the verification set are highly consistent in time, and further make the weight curve and the verification curve highly consistent in time. The weight curve and the verification curve are plotted in the same graph to obtain a comparison curve graph. It should be noted that the weight curve and the verification curve are not fused. At this time, the time series of the weight curve and the time series of the verification curve are the same. Then the time series of the comparison curve graph is actually a common time series. Since the data in the weight result and the verification result correspond to two types of normal data and abnormal data, in order to comprehensively analyze the difference between the data and the corresponding data at the same time in history, the normal data is divided into one or more intervals based on the common time series, and the abnormal data is divided into one or more intervals. The weight trend and the verification trend in the curve interval are extracted, and the weight trend and the verification trend are compared to obtain the difference between the data and the corresponding data at the same time in history, that is, the time period when the weight trend direction is the same as the verification trend direction and the time period when the weight trend direction is different from the verification trend direction. Then, the time period when the weight trend direction is different from the verification trend direction, the weight set, and the verification set are sorted out to obtain the verification difference. At this time, the verification difference can be used to quantify the performance of the anomaly evaluation model, and the accuracy of the anomaly evaluation model in using the monitoring data to judge whether there are potential safety hazards in the structural state of the wind turbine tower can be obtained. Further, the threshold α in the constraint condition can be adjusted to effectively improve the accuracy of the anomaly evaluation model.

[0034] Obtain real-time monitoring data of the top of the wind turbine tower, and input the real-time monitoring data into an accurate evaluation model to obtain an evaluation result; correspondingly, the structural state monitoring system of the wind turbine tower performs an alarm action in response to the evaluation result; Specifically, extract the real-time monitoring data determined to be abnormal in the evaluation result, and determine the abnormal wind turbine tower based on the real-time monitoring data; Extract the operation and maintenance information and operation and maintenance phone numbers of the corresponding abnormal wind turbine tower from the operation and maintenance information database, and convert the operation and maintenance information into a text message and send it to the operation and maintenance phone number.

[0035] In this embodiment, the evaluation result at least includes monitoring data and the status of the monitoring data. When the vibration intensity of the corresponding wind turbine tower calculated by the accurate evaluation model based on the real-time monitoring data is outside the corresponding constraint conditions, that is, the absolute value of the vibration intensity is greater than the absolute value of the corresponding threshold α, the corresponding monitoring data is marked as abnormal, otherwise it is marked as normal.

[0036] On the other hand, as Figure 2 shown, another technical solution provided in the embodiment of the present invention is a structural state monitoring system for a wind turbine tower, including: a data acquisition module, a principal component module, an evaluation module, and an alarm module; The data acquisition module is used to collect and preprocess the monitoring data at the top of the wind turbine tower to obtain a weight set, a validation set, and real-time monitoring data, transmit the weight set to the principal component module, and transmit the real-time monitoring data and the validation set to the evaluation module; The principal component module performs principal component analysis on the weight set uploaded by the data acquisition module to obtain data contribution degrees, and transmits the data contribution degrees to the evaluation module; The evaluation module constructs an abnormal evaluation model based on the data contribution degrees uploaded by the principal component module, validates the abnormal evaluation model based on the validation set uploaded by the data acquisition module to obtain an accurate evaluation model, and transmits the evaluation result obtained by inputting the real-time monitoring data uploaded by the data acquisition module into the accurate evaluation model to the alarm module; The alarm module performs an alarm action based on the evaluation result uploaded by the evaluation module.

[0037] This embodiment at least has the following substantial effects: (1) In this embodiment, principal component analysis is performed on the historical monitoring data of the wind turbine tower to obtain the data contribution degree. Considering that the tilt angle, vibration displacement, and vibration acceleration in the historical monitoring data are all related to the vibration intensity at the top of the wind turbine tower, and the vibration intensity can intuitively reflect the overall structural stability of the wind turbine tower, the data contribution degree can represent the correlation degree between different types of data in the monitoring data and the vibration intensity. Furthermore, an abnormal evaluation model can be established based on this correlation degree, enabling the abnormal evaluation model to directly use data such as tilt angle, vibration displacement, and vibration acceleration to evaluate the vibration intensity of the corresponding wind turbine tower, thereby realizing the monitoring of potential safety hazards of the wind turbine tower; (2) In this embodiment, the verification set is used to verify the abnormal evaluation model to obtain an accurate evaluation model. By taking the verification set as a benchmark to adjust the constraint conditions of the abnormal evaluation model, the performance of the abnormal evaluation model is guaranteed. Then, the real-time monitoring data is input into the accurate evaluation model to obtain the evaluation result, which at least includes the monitoring data and the status of the monitoring data. That is, if there are potential safety hazards in the wind turbine tower corresponding to the monitoring data, the monitoring data is abnormal monitoring data; if there are no potential safety hazards in the wind turbine tower corresponding to the monitoring data, the monitoring data is normal monitoring data. Finally, the corresponding wind turbine tower structure state monitoring system executes an alarm action in response to the evaluation result. By using the data contribution degree, the commonality of the massive monitoring data is used as the basis for the accurate evaluation model, and thus the accurate evaluation model can be used to quickly process the massive monitoring data, effectively improving the real-time performance of wind turbine tower monitoring.

[0038] The above specific implementation manners are the preferred implementation manners of the present invention, which do not limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to this specific implementation manner. All equivalent changes made according to the shape, structure, and method of the present invention are within the protection scope of the present invention.

Claims

1. A method for monitoring the structural status of a wind power tower, characterized in that: The following steps are involved: Obtain historical monitoring data from the top of the wind turbine tower and perform preprocessing to obtain a weight set and a validation set; Perform principal component analysis on the weighted data to obtain data contribution; An anomaly assessment model is constructed based on data contribution, and the anomaly assessment model is verified using the validation set to obtain an accurate assessment model; Obtain real-time monitoring data from the top of the wind tower and input the real-time monitoring data into an accurate assessment model to obtain the assessment results; The corresponding wind turbine tower structure condition monitoring system performs an alarm action in response to the evaluation result.

2. A method for monitoring the structural status of a wind power tower according to claim 1, characterized in that: The specific process of preprocessing the historical monitoring data on the top of the wind power tower to obtain the weight set and the verification set is as follows: Counting the abnormal values ​​and missing values ​​in the historical monitoring data at the top of the wind tower, and deleting the abnormal values ​​to obtain a missing data set; Fill the missing data in the missing data set based on the interpolation method to obtain a complete data set, and normalize the data in the complete data set based on the normalization criterion to obtain a standard data set; The standard data set is divided based on the preset weight-verification ratio to obtain the weight set and the verification set.

3. A method for monitoring the structural status of a wind power tower according to claim 1, characterized in that: The specific process of performing principal component analysis on the weighted data to obtain data contribution is as follows: Decentralizing the data based on the mean of the data in the weighted data set to obtain decentralized data; Construct a feature matrix based on the decentralized data, and transform the feature matrix into a covariance matrix based on the covariance matrix transformation criterion; Decompose the covariance matrix based on the eigenvalue solution criterion to obtain the principal component direction and variance contribution rate; The principal components are determined based on the variance contribution rate and the principal component direction, and the contribution of the data in the weighted data set is calculated based on the principal components to obtain the data contribution.

4. A method for monitoring the structural status of a wind power tower according to claim 3, characterized in that: The mathematical model of the covariance matrix conversion criterion is specifically: In the formula, C is the covariance matrix, X is the feature matrix, and X T is the transposed matrix of the feature matrix, and n is the number of samples corresponding to the decentralized data.

5. A method for monitoring the structural status of a wind power tower according to claim 3, characterized in that: The specific process of determining the principal component based on the variance contribution rate and the principal component direction is as follows: The variance contribution rates of all the undetermined principal components in the principal component direction are sorted by size, and the cumulative variance contribution rate is accumulated starting from the largest variance contribution rate; When the cumulative variance contribution rate is greater than or equal to a preset contribution rate threshold, the accumulation of the variance contribution rate is stopped, and the undetermined principal component corresponding to the cumulative variance contribution rate is marked as the principal component.

6. A method for monitoring the structural status of a wind power tower according to claim 3, characterized in that: The contribution formula corresponding to the data contribution is specifically: Where Y i is the data contribution of the i-th decentralized data, α ij is the load of the i-th decentralized data on the j-th principal component, y j is the variance contribution rate of the jth principal component, and p is the total number of principal components.

7. A method for monitoring the structural status of a wind power tower according to claim 1, characterized in that: The specific process of constructing an anomaly assessment model based on data contribution and using a validation set to validate the anomaly assessment model to obtain an accurate assessment model is as follows: The weight of the data in the corresponding weighted data set is set based on the data contribution, and the data features of the data in the weighted data set are extracted; Constructing an anomaly assessment model based on the weights and data features, and inputting the weight set into the anomaly assessment model to obtain a weight result; Compare and analyze the weight results with the validation set to obtain the validation difference, and adjust the anomaly assessment model based on the validation difference; The abnormal evaluation model that has completed the adjustment is marked as the accurate evaluation model.

8. A method for monitoring the structural status of a wind power tower according to claim 7, characterized in that: The specific process of comparing and analyzing the weight results with the validation set to obtain the validation difference is as follows: Based on the time series of the weight set corresponding to the weight result, the validation set is aligned to obtain a common time series, and the aligned validation set is input into the anomaly assessment model to obtain the validation result; The data corresponding to the weight result is plotted as a weight curve graph, and the data corresponding to the verification result is plotted as a verification curve graph; Integrate the weight curve graph and the verification curve graph to obtain a comparison curve graph; Divide the curve interval of the comparison curve graph based on the common time series, and extract the weight trend and verification trend in the curve interval; The weight trend is compared with the verification trend, and the time periods, weight sets and verification sets with different weight trend trends are sorted out to obtain the verification differences.

9. A method for monitoring the structural status of a wind power tower according to claim 1, characterized in that: The specific process of the corresponding wind power tower structure status monitoring system executing the alarm action in response to the evaluation result is: Extracting the real-time monitoring data determined to be abnormal in the evaluation results, and determining the abnormal wind power tower based on the real-time monitoring data; The operation and maintenance information and operation and maintenance telephone number corresponding to the abnormal wind power tower are extracted based on the operation and maintenance information database, and the operation and maintenance information is converted into a text message and sent to the operation and maintenance telephone number.

10. A wind power tower structure status monitoring system, applicable to a wind power tower structure status monitoring method as claimed in any one of claims 1 to 9, characterized in that: It includes: data acquisition module, main component module, evaluation module and alarm module; The data acquisition module is used to collect and pre-process the monitoring data on the top of the wind power tower to obtain a weight set, a verification set, and real-time monitoring data, transmit the weight set to the principal component module, and transmit the real-time monitoring data and the verification set to the evaluation module; The principal component module performs principal component analysis based on the weight set uploaded by the data acquisition module to obtain data contribution, and transmits the data contribution to the evaluation module; The evaluation module constructs an abnormal evaluation model based on the data contribution uploaded by the principal component module, verifies the abnormal evaluation model based on the verification set uploaded by the data acquisition module to obtain an accurate evaluation model, and inputs the real-time monitoring data uploaded by the data acquisition module into the accurate evaluation model to obtain an evaluation result which is transmitted to the alarm module; The alarm module performs an alarm action based on the evaluation result uploaded by the evaluation module.

Images