A method for monitoring the state of a blade

Abstract

The present invention provides a method for monitoring the blade state. Under the condition that the blade is not damaged, the blade reference vibration data and reference operation monitoring data at multiple cross-sections along the direction from the blade root to the blade tip on the same blade are collected to obtain the reference frequency-domain covariance. Then, a memory matrix is constructed based on the reference operation monitoring data, and the reference similarity threshold between any two cross-sections is calculated. Under the actual working condition of the blade, the actual frequency-domain covariance, actual operation monitoring data, and memory matrix of two cross-sections of the same blade are collected, and the actual similarity value between any two cross-sections is calculated. The actual similarity value between the two cross-sections is compared with the reference similarity threshold to determine the damage condition of the blade. The present disclosure can significantly reduce the usage amount of sensors, and the damage position of the blade can be monitored in real time based on the constructed similarity damage model.

Description

Technical Field

[0001] The present invention relates to the field of measurement technologies, and particularly to a method for monitoring the state of a blade. Background Art

[0002] In recent years, with the rapid development of large wind turbines (above 5 MW), especially the large-scale installation of 10 MW offshore wind turbines, blade fracture and falling accidents occur frequently. Currently, the main online monitoring technologies for blade damage of large wind turbines are as follows: 1) Damage monitoring based on the change of blade strain. However, this monitoring method is limited to the damage of local materials and cannot effectively detect damage far from the sensor installation position. Although increasing the number of sensors used can improve the measurement effect to a certain extent, the cost of installing strain sensors over a large area is unacceptable. 2) Damage monitoring based on acoustic emission sensors. The theoretical basis of acoustic emission technology is that when a composite material is damaged, the damaged part will generate high-frequency elastic waves at the moment of material failure, which is the so-called acoustic emission. Acoustic emission sensors monitor whether such elastic waves appear in the material as the basis for blade damage monitoring. Since the acoustic emission signal appears briefly during local material damage and then the load redistributes to the surrounding composite materials that have not yet been damaged, the acoustic emission signal disappears, and there is no acoustic emission signal between the acoustic emission signals generated by adjacent local material damages, even though the damage has occurred at this time. In addition, when the blade enters the pitch adjustment mode during the operation of the wind turbine, the movement of the pitch adjustment mechanism will generate a large amount of noise, which propagates in the blade material. At this time, the load of the wind turbine blade is often in a large state, that is, when the probability of blade material damage is relatively high, the noise of the pitch adjustment mechanism affects the detection of damage signals by acoustic emission sensors. Therefore, the reliability of using acoustic emission signals to evaluate blade damage cannot be guaranteed. 3) Monitoring the change of the natural frequency of the blade by vibration acceleration sensors as the basis for blade damage monitoring. However, many research reports show that during the bench fatigue test of full-scale blades, the natural frequency of the blade is collected by the method of single-point loading and release stage by stage, and the results show that the change of the natural frequency of the blade is so small that it can be ignored, although the blade has suffered different degrees of damage. For wind turbines in actual operation, the excitation of the natural frequency of the blade cannot be carried out by artificial loading methods. And relying on the natural excitation of the wind cannot reliably obtain the natural frequency of the blade.

[0003] Therefore, how to monitor the health state of the blade, especially monitor blade damage, at low cost, accurately and persistently is an urgent problem to be solved at present. Summary of the Invention

[0004] In order to overcome at least one of the many problems in the related technologies, the present invention provides a method for monitoring the state of a blade.

[0005] Among them, the method for monitoring the state of the blade is applied to a wind power generation device and includes:

[0006] Collect the blade reference vibration data at multiple cross-sections along the direction from the blade root to the blade tip on the same blade without damage to the blade, and obtain the reference frequency-domain covariance between any two of the cross-sections based on the reference vibration data;

[0007] Collect the reference operation monitoring data of the wind power generation equipment without damage to the blade;

[0008] Construct a memory matrix according to the reference frequency-domain covariance and the reference operation monitoring data, and then calculate the reference similarity threshold between any two cross-sections;

[0009] Collect the actual vibration data of the same blade at multiple cross-sections along the direction from the blade root to the blade tip under the actual working condition of the blade, and obtain the actual frequency-domain covariance between any two of the cross-sections based on the actual vibration data;

[0010] Collect the actual operation monitoring data of the wind power generation equipment under the actual working condition of the blade;

[0011] Calculate the actual similarity value between any two cross-sections according to the actual frequency-domain covariance, the actual operation monitoring data and the memory matrix;

[0012] Compare the actual similarity value between two cross-sections with the reference similarity threshold to determine the damage condition of the blade.

[0013] In some optional embodiments, without damage to the blade and within a first preset time period, for any blade m, any cross-section on it SEC t of the k line of the reference amplitude spectrum of the BL m _SEC t _AMP k is:

[0014] ,

[0015] wherein, k is the serial number of the line of the reference amplitude spectrum of the k th item, , N is the total length of the line sequence of the reference amplitude spectrum; is the SEC t th item of the reference vibration data sequence collected at the cross-section n .

[0016] In some alternative embodiments, the reference frequency-domain covariance between any two of the cross-sections is obtained according to the following formula:

[0017] ,

[0018] where is the reference frequency-domain covariance between cross-section and cross-section ; is the spectral line of the th k item of the reference amplitude spectrum of cross-section ; is the average value of the spectral lines of the reference amplitude spectrum of cross-section ; is the spectral line of the k th item of the reference amplitude spectrum of cross-section ;

[0019]

[0020] In some alternative embodiments, constructing a memory matrix based on the reference frequency-domain covariance and the reference operation monitoring data includes: Creating a reference data set by using the reference frequency-domain covariance and the reference operation monitoring data;

[0021] Downsampling the reference data set to obtain a downsampled reference frequency-domain covariance and downsampled reference operation monitoring data;

[0022] Combining the downsampled reference frequency-domain covariance and the downsampled reference operation monitoring data to form the memory matrix.

[0023] In some alternative embodiments, calculating the reference similarity threshold between any two cross-sections includes:

[0024] Removing the downsampled reference frequency-domain covariance and the downsampled reference operation monitoring data from the reference data set to obtain a training data set;

[0025] Selecting the data corresponding to any moment T from the training data set as the observation vector ;

[0026] Calculating the estimated vector based on the observation vector ;

[0027] Calculating the similarity between the observation vector corresponding to any moment T and the estimated vector ;

[0028] Repeat the above steps to calculate the similarity corresponding to each moment in the training dataset to construct a similarity set , and based on the said similarity set obtain a reference similarity threshold

[0029] In some alternative embodiments, the calculating of the estimated vector based on the said observation vector is performed according to the following formula

[0030] ,

[0031] where D is the said memory matrix

[0032] In some alternative embodiments, the similarity between the observation vector corresponding to any moment T and the estimated vector is calculated according to the following formula :

[0033] sim = 1 1+ ∑ h = 1 H [ V est (h)- V obs (h)] 2 ,

[0034] where ; H is the total number of monitoring items of the reference operation monitoring data of the wind power generation equipment

[0035] In some alternative embodiments, the calculating of the actual similarity value between any two cross-sections according to the said actual frequency domain covariance, the said actual operation monitoring data and the memory matrix includes

[0036] constructing an actual observation vector according to the actual frequency domain covariance and the actual operation monitoring data ;

[0037] calculating the actual estimated vector between any cross-sections according to the actual observation vector and the said memory matrix ;

[0038] calculating the actual similarity value between any two cross-sections according to the actual observation vector and the actual estimated vector

[0039] In some alternative embodiments, the reference similarity threshold includes a damage warning threshold

[0040] If the actual similarity value is less than the damage warning threshold, it indicates that the blade is damaged

[0041] ​In some alternative embodiments, the damage warning threshold is calculated by the following formula:

[0042] ,

[0043] Wherein, is the damage warning threshold, is the warning threshold coefficient; and are the mean and standard deviation of the similarity set respectively.

[0044] The technical solution of the present invention has the following advantages or beneficial effects:

[0045] The present disclosure uses the multi-section vibration data of the wind turbine blade and the monitoring data of the wind power generation equipment as the blade damage monitoring indicators. Therefore, only several acceleration sensors need to be installed along the length direction of each blade, and the number of sensors does not need to cover the blade surface in a large area. By collecting the operation data of the wind power generation equipment, it is possible to effectively monitor whether the blade is damaged and locate the blade damage position. Correspondingly, the accuracy of damage location depends on the number of sensors. The more the number, the more accurate the damage position monitoring. In addition, once the damage occurs, this actual similarity value will always remain and will deviate more and more from the warning line as the damage degree increases, rather than having no obvious change like the natural frequency. In summary, the present disclosure can greatly reduce the usage amount of sensors, and based on the constructed similarity damage model, it is possible to monitor the blade damage position in real time. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The drawings are used to better understand the present invention and do not constitute an improper limitation to the present invention. Among them:

[0047] Figure 1 is a schematic diagram of the position of the blade monitoring section and the installation position of the sensor according to an embodiment of the present invention;

[0048] Figure 2 is a schematic diagram of the change of the damage factor of the blade in the non-damaged state with the monitoring time according to an embodiment of the present invention;

[0049] Figure 3 is a schematic diagram of the change of the damage factor of the blade in the damaged state with the monitoring time according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0050] The following describes exemplary embodiments of the present invention with reference to the accompanying drawings. Various details of the embodiments of the present invention are included to facilitate understanding, and they should be considered merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, descriptions of well-known functions and structures are omitted in the following description for clarity and conciseness.

[0051] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms "a", "the", and "said" used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0052] It should be understood that although the terms first, second, third, etc. may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of this application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "when" or "while" or "in response to determining".

[0053] As described in the background art section, in order to solve at least one of the many problems in the prior art, the present disclosure provides a method for monitoring the blade state of a wind power generation device. The method includes: collecting blade reference vibration data at multiple cross-sections along the root-to-tip direction of the same blade when the blade is undamaged, and obtaining the reference frequency-domain covariance between any two of the cross-sections based on the reference vibration data; collecting reference operation monitoring data of the wind power generation device when the blade is undamaged; constructing a memory matrix according to the reference frequency-domain covariance and the reference operation monitoring data, and then calculating the reference similarity threshold between any two cross-sections; collecting actual vibration data of the same blade at multiple cross-sections along the root-to-tip direction when the blade is actually working, and obtaining the actual frequency-domain covariance between any two of the cross-sections based on the actual vibration data; collecting actual operation monitoring data of the wind power generation device when the blade is actually working; calculating the actual similarity value between any two cross-sections according to the actual frequency-domain covariance, the actual operation monitoring data, and the memory matrix; comparing the actual similarity value between two cross-sections with the reference similarity threshold to determine the damage condition of the blade.

[0054] As Figure 1In the illustrated embodiment, for any blade of a wind power generation device, a plurality of cross-sections are arbitrarily selected from the blade root to the blade tip direction, and sensors are installed for condition monitoring. For example, for the same blade, the number of monitoring cross-sections can be set to 2, 3, 4, 5, etc. The distances between the monitoring surfaces can be equal or unequal. In actual use, the number of monitoring cross-sections can be flexibly adjusted according to needs, and no specific limitation is made here. It can be understood that according to the monitoring method of the present disclosure, the greater the density of the set monitoring cross-sections, the more accurately the damage position of the blade can be located. Of course, for cost considerations, the solutions provided in some embodiments of the present invention can still ensure the positioning accuracy of the damage position and the accuracy of blade damage monitoring while appropriately reducing the number of monitoring cross-sections. Figure 1 In the illustrated embodiment, only 3 monitoring cross-sections are provided, namely cross-section A (hereinafter denoted as surface ), cross-section B (hereinafter denoted as surface ), and cross-section C (hereinafter denoted as surface ). And sensors are provided at the corresponding cross-sections (such as Figure 1as shown by the black square in). The sensor is used to measure the vibration data of the blade, including but not limited to an acceleration sensor, etc. Of course, other measuring devices that can capture the blade response signal can also be selected. After the measuring device is installed at the corresponding monitoring section of the blade, vibration monitoring can be carried out in the flapping direction of the blade. Among them, the flapping direction of the blade refers to the bending vibration direction of the blade in the direction perpendicular to the rotation plane. Specifically, according to the definition of the blade coordinate system, the flapping direction of the blade refers to the axis direction that coincides with the pitch axis of the blade, is perpendicular to the rotation plane, and for an upwind wind turbine, the positive direction points to the tower direction. When monitoring blade damage, first a reference needs to be selected, and it is judged whether there is damage to the blade based on the difference between the actual operating state of the blade and the reference operating state. For this purpose, in some embodiments of the present invention, when the blade is undamaged, the reference vibration data of the blade at multiple sections along the root-to-tip direction of the same blade is collected, and the reference frequency-domain covariance between any two of the sections is obtained based on the reference vibration data; when the blade is undamaged, the reference operation monitoring data of the wind power generation equipment is collected; a memory matrix is constructed according to the reference frequency-domain covariance and the reference operation monitoring data, and then the reference similarity threshold between any two sections is calculated; finally, the reference similarity threshold is used as the judgment criterion for whether there is damage to the blade. After the reference is selected, the present disclosure selects to monitor the actual vibration data during the operation of the blade and the actual operation monitoring items of the wind power generation equipment. The monitoring items include but are not limited to impeller speed, wind speed, active power, pitch angle, yaw angle, wind direction angle, and front and rear acceleration of the nacelle, etc. Therefore, when the blade is actually working, the actual vibration data of the blade at multiple sections along the root-to-tip direction of the same blade is collected, and the actual frequency-domain covariance between any two of the sections is obtained based on the actual vibration data; when the blade is actually working, the actual operation monitoring data of the wind power generation equipment is collected; according to the actual frequency-domain covariance, the actual operation monitoring data, and the memory matrix, the actual similarity value between any two sections is calculated. Finally, for the same blade, the damage condition of the blade is determined by comparing the difference between the actual similarity value between two sections and the reference similarity threshold. For example, for blade m, by comparing the actual similarity value between section A and section B on it with the reference similarity threshold between section A and section B, it can be determined whether there is damage between section A and section B of the blade. By monitoring the similarity between different sections, the location where the damage occurs can be determined.

[0055] As can be seen from the above description, the present disclosure uses the multi-section vibration data of the fan blade and the monitoring data of the wind power generation equipment as the blade damage monitoring indicators. Therefore, only a number of acceleration sensors need to be installed along the length direction of each blade, and the number of sensors does not need to cover the blade surface in a large area, and collecting the operation data of the wind power generation equipment can effectively monitor whether the blade is damaged and locate the blade damage position. Correspondingly, the accuracy of damage location depends on the number of sensors. The more the number, the more accurate the damage location monitoring. In addition, once the damage occurs, this actual similarity value will always remain and will deviate further from the warning line as the damage degree increases, rather than having no obvious change like the natural frequency. In summary, the present disclosure can greatly reduce the usage amount of sensors, and based on the constructed similarity damage model, the blade damage position can be monitored in real time.

[0056] In an optional embodiment, in the case where the blade is not damaged and within the first preset time period, for any blade m, any cross-section on it of the th spectral line of the reference amplitude spectrum is:

[0057] ,

[0058] wherein, is the serial number of the spectral line of the th reference amplitude spectrum, , is the total length of the spectral line sequence of the reference amplitude spectrum; is the th item of the reference vibration data sequence collected at the cross-section .

[0059] Generally, a wind power generation equipment has 3 blades, so m can take any number from 1 to 3. As shown in the Figure 1 embodiment, a total of three cross-sections are selected, namely , and . Therefore can be , and SEC c any one of them.

[0060] It should be noted that the above reference vibration data sequence can be the vibration data sequence collected within the first preset time period, such as the sequence composed of monitoring data such as frequency and amplitude in the time domain.

[0061] In some optional embodiments, the reference frequency domain covariance between any two of the cross-sections is obtained according to the following formula:

[0062] ,

[0063] wherein, is the reference frequency-domain covariance between cross-section and cross-section ; is the spectral line of the th term of the reference amplitude spectrum of cross-section ; is the average value of the spectral lines of the reference amplitude spectrum of cross-section ; is the th spectral line of the reference amplitude spectrum of cross-section ;

[0064] In the above embodiments, the reference frequency-domain covariance between cross-section A and cross-section B is calculated as an example. When the corresponding data of cross-section B and cross-section C are substituted into the above formula, the reference frequency-domain covariance between cross-section B and cross-section C can be calculated. When more cross-sections are selected, when calculating the reference frequency-domain covariance between any two cross-sections, only the data of the corresponding cross-sections need to be substituted into the above formula.

[0065] In some alternative embodiments, constructing a memory matrix based on the reference frequency-domain covariance and the reference operation monitoring data includes:

[0066] Creating a reference data set using the reference frequency-domain covariance and the reference operation monitoring data;

[0067] Downsampling the reference data set to obtain downsampled reference frequency-domain covariance and downsampled reference operation monitoring data;

[0068] Combining the downsampled reference frequency-domain covariance and the downsampled reference operation monitoring data to form the memory matrix.

[0069] In the present invention, under the condition of no damage, the operation data of the blade and the wind power generation equipment are jointly used as an evaluation criterion to determine whether there is damage during the actual operation of the blade, rather than simply comparing the data of the blade itself under the condition of no damage and the data under the actual operation condition to determine whether there is damage during the actual operation of the blade. In other words, in some embodiments of the present invention, the operation data of the wind power generation equipment is further introduced to jointly determine the damage state of the blade. Thereby effectively eliminating random errors and improving the accuracy and reliability of damage monitoring. For this reason, the damage judgment model constructed in some embodiments of the present invention needs to construct a memory matrix.

[0070] Specifically, to construct the memory matrix, a reference data set, denoted as B, needs to be created first. The reference data set B includes two parts. The first part is the reference frequency domain covariance in the non-damaged state of the blade described above, and the second part is the SCADA (Supervisory Control And Data Acquisition) data of the wind power generation equipment during the same period, that is, the data acquisition and monitoring control system data. The data includes, but is not limited to, monitoring items such as impeller speed, wind speed, active power, pitch angle, yaw angle, wind direction angle, and front and rear acceleration of the nacelle. In some preferred embodiments, the monitoring item data needs to go through a data cleaning process to remove outliers and noise, etc., to improve the accuracy of blade damage monitoring. In some other embodiments, due to the inconsistent dimensions of various data in the reference data set, normalization processing is also required to unify all data ranges so that they are between [0, 1], to improve data quality and ensure the accuracy of subsequent calculations.

[0071] After completing the above data processing, downsampling is performed on the reference data set B to obtain the downsampled reference frequency domain covariance and the downsampled reference operation monitoring data . Among them, the downsampling method can adopt the commonly used methods in the prior art; for example, select some data from the original data, and the time intervals corresponding to the selected data are equal.

[0072] Finally, combine the downsampled reference frequency domain covariance and the downsampled reference operation monitoring data to form the memory matrix D, and its expression is as follows:

[0073] ,

[0074] As Figure 1 shown in the embodiment, there are a total of 3 cross-sections, that is , and . There is a memory matrix corresponding between and ; There is a memory matrix

[0075] ,

[0076] ,

[0077] Among them, represents the data corresponding to the th monitoring item in ; H represents the total number of monitoring items; represents the time point of monitoring, , that is, the data at G moments are monitored in total.

[0078] In some optional embodiments, calculating the reference similarity threshold between any two cross-sections includes:

[0079] Removing the downsampled reference frequency-domain covariance and downsampled reference operation monitoring data from the reference data set to obtain a training data set;

[0080] Selecting the data corresponding to any moment T from the training data set as the observation vector ;

[0081] Based on the observation vector Calculating the estimated vector ;

[0082] Calculating the similarity between the observation vector corresponding to any moment T and the estimated vector ; ;

[0083] Repeating the above steps to calculate the similarity corresponding to each moment in the training data set to construct a similarity set , and based on the similarity set Obtaining the reference similarity threshold.

[0084] As described above, in some embodiments of the present invention, the determination of blade damage is based on the operating parameters of the blade and the operating parameters of the wind power generation equipment together. Therefore, it is necessary to construct a reference similarity threshold between two cross-sections based on these data, and then judge the difference between the actual similarity of the blade and the reference similarity threshold to determine whether the blade is damaged.

[0085] Specifically, first, remove the memory matrix from the reference data set to obtain a training data set , and then take the data corresponding to any moment T in the training data set as the observation vector , and calculate the estimated vector . Among them, the estimated vector is calculated by the following formula:

[0086] ,

[0087] Among them, is the memory matrix and the training data set Observation vector and the weight vector of

[0088] Taking the three cross-sections A, B, and C corresponding to the Figure 1 blade described above as an example; the estimated vector between cross-sections A and B and the estimated vector between cross-sections B and C are calculated respectively according to the following formulas:

[0089] ,

[0090] ,

[0091] where and are the observation vectors between cross-sections A and B and between cross-sections B and C respectively.

[0092] Furthermore, calculate the similarity between the observation vector corresponding to any moment T described above and the estimated vector , and the specific calculation formula is as follows:

[0093] sim = 1 1+ ∑ h = 1 H [ V est (h)- V obs (h)] 2 ,

[0094] where ; H is the total number of monitoring items of the reference operation monitoring data of the wind power generation equipment.

[0095] Taking the three cross-sections A, B, and C corresponding to the Figure 1 blade described above as an example; the similarity between cross-sections A and B and the similarity between cross-sections B and C are calculated respectively according to the following formulas:

[0096] sim A&B = 1 1+ ∑ h = 1 H [ V est_A&B (h)- V obs_A&B (h)] 2 ,

[0097] sim B &C = 1 1+ ∑ h = 1 H [ V est_B&C (h)- V obs_B&C (h)] 2 ,

[0098] Similarly, calculate the similarity at each moment in the training data set , and form a similarity set with all the similarities. Finally, obtain the reference similarity threshold based on the similarity set described above. Preferably, the reference similarity threshold is calculated according to the following formula:

[0099] ,

[0100] ,

[0101] wherein, and are the similarity warning threshold and the alarm threshold respectively; and are the warning threshold coefficient and the alarm threshold coefficient respectively, and both are determined by the actual operating conditions of the blade, and is less than ; and are the mean value and the standard deviation of the similarity set respectively.

[0102] For Figure 1 among the three cross-sections, there are two sets of reference similarity thresholds corresponding to the cross-sections A and B and the cross-sections B and C respectively. Each set of similarity thresholds includes a similarity warning threshold and an alarm threshold. Therefore, there are a total of 4 similarity thresholds, and the calculation formula is as follows:

[0103] ,

[0104] ,

[0105] ,

[0106] ,

[0107] wherein, the subscript corresponding parameter represents the data associated with the cross-sections A and B; the subscript corresponding parameter represents the data associated with the cross-sections A and B.

[0108] In some alternative embodiments, calculating the actual similarity value between any two cross-sections according to the actual frequency domain covariance, the actual operation monitoring data, and the memory matrix includes:

[0109] Constructing an actual observation vector according to the actual frequency domain covariance and the actual operation monitoring data ;

[0110] Specifically, during the actual operation of the wind power generation equipment, several consecutive time SCADA data (i.e., the actual operation monitoring data) are taken, and the frequency domain covariance between any two cross-sections of any blade corresponding to the same time is calculated. Then, using the method of constructing the observation vector described above, the actual operation data of the blade is used to construct the actual observation vector . In some preferred embodiments, data cleaning and normalization are also performed to unify all data ranges to between [0, 1].

[0111] Then, based on the actual observation vector and the memory matrix mentioned above, calculate the actual estimated vector between any two cross-sections . The specific calculation formula is as follows:

[0112] ,

[0113] ,

[0114] where the subscript corresponding parameter represents the data associated with cross-sections A and B; the subscript corresponding parameter represents the data associated with cross-sections A and B.

[0115] Based on the actual observation vector and the actual estimated vector calculate the actual similarity value between any two cross-sections . The specific calculation formula is as follows:

[0116] sim1 A&B = 1 1+ ∑ h = 1 H [ V1 est_A&B (h)- V1 obs_A&B (h)] 2 ,

[0117] sim1 B &C = 1 1+ ∑ h = 1 H [ V1 est_B&C (h)- V1 obs_B&C (h)] 2 ,

[0118] where the subscript corresponding parameter represents the data associated with cross-sections A and B; the subscript corresponding parameter represents the data associated with cross-sections A and B. ; H is the total number of monitoring items of the reference operation monitoring data of the wind power generation equipment.

[0119] In some alternative embodiments, the reference similarity threshold includes a damage warning threshold: if the actual similarity value is less than the damage warning threshold, it indicates that the blade is damaged.

[0120] Taking the monitoring of 3 blades of a wind power generation equipment as an example. Taking the actual similarity between the observation vector and the estimated vector during actual operation as the blade damage factor; taking the reference similarity threshold and as the blade damage early warning threshold and the warning threshold respectively.

[0121] Taking cross-section and cross-section Taking it as an example (the evaluation methods between other cross-sections are the same), the damage condition of the blade can be judged by comparing the blade damage factor with the blade damage warning and alarm thresholds. Specifically as follows:

[0122] ① , it indicates that there is no risk of structural damage in the area between the cross-section and the cross-section of the blade at the current moment;

[0123] ② If , it indicates that there is a certain risk of structural damage in the area between the cross-section and the cross-section of the blade at the current moment. A warning should be triggered and the monitoring should be strengthened;

[0124] ③ , it indicates that there is a relatively large risk of structural damage in the area between the cross-section and the cross-section of the blade at the current moment. At this time, an alarm should be triggered immediately and professional personnel should be arranged to conduct a safety inspection on the blade.

[0125] As can be seen from the above description, in the embodiment shown in Figure 2 , when the blade is undamaged, the damage factor only fluctuates slightly over time. In Figure 3 , when the blade is damaged for the first time, its damage factor will fluctuate downward. And as time goes by, the damage will gradually become serious, and then the damage factor will change violently. That is to say, once the damage appears, the damage factor will change accordingly and will not disappear, which is convenient for long-term tracking and evaluation of the blade damage state. The method of the present disclosure does not need to judge the blade abnormality according to the change of the natural frequency of the blade, and does not need to know the value of the natural frequency of the blade in advance. The sensitivity is greatly improved compared with the measurement of the change of the natural frequency. It can be seen that the method of the present disclosure can not only detect the blade abnormality, but also determine the specific location where the damage appears according to the change of the damage factor between adjacent cross-sections, that is, determine between which two adjacent cross-sections the damage appears.

[0126] The above specific implementation manners do not constitute a limitation to the protection scope of the present invention. Those skilled in the art will easily think of other implementation manners of the present disclosure after considering the specification and practicing the technical solutions disclosed in this application. This application aims to cover any variations, uses or adaptations of the present disclosure, and these variations, uses or adaptations follow the general principles of the present disclosure and include the common general knowledge or conventional technical means in the technical field not disclosed in the present disclosure. The specification and examples are only regarded as exemplary, and the true scope and spirit of the present disclosure are pointed out by the following claims.

[0127] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A blade status monitoring method, applied to wind power generation equipment, characterized in that: Collecting blade reference vibration data at multiple sections along the direction from the blade root to the blade tip on the same blade when the blade is not damaged, and obtaining a reference frequency domain covariance between any two of the sections based on the reference vibration data; Collecting the benchmark operation monitoring data of the wind power generation equipment when the blades are not damaged; Constructing a memory matrix based on the benchmark frequency domain covariance and the benchmark operation monitoring data, and then calculating a benchmark similarity threshold between any two sections; Under actual working conditions of the blade, actual vibration data of the blade at a plurality of the sections along the direction from the blade root to the blade tip on the same blade are collected, and actual frequency domain covariance between any two of the sections is obtained based on the actual vibration data; Collecting actual operation monitoring data of the wind power generation equipment when the blades are actually working; Calculate the actual similarity value between any two cross sections according to the actual frequency domain covariance, the actual operation monitoring data and the memory matrix; The actual similarity value between the two cross sections is compared with a reference similarity threshold to determine the damage condition of the blade.

2. The blade status monitoring method according to claim 1, characterized in that: When the blade is not damaged and within the first preset time period, for any blade m, any section on it No. Spectral lines of the reference amplitude spectrum for: , in, It is The number of the spectral line of the reference amplitude spectrum, , is the total length of the spectral line sequence of the reference amplitude spectrum; It is in the cross section The first part of the baseline vibration data sequence collected at item.

3. The blade status monitoring method according to claim 1, characterized in that: The reference frequency domain covariance between any two sections is obtained according to the following formula: ], in, It is a cross section and cross section The reference frequency domain covariance between It is a cross section No. The spectral lines of the reference amplitude spectrum, It is a cross section The average value of the spectral lines of the reference amplitude spectrum; It is a cross section No. The spectral lines of the reference amplitude spectrum, It is a cross section The average value of the spectral lines of the reference amplitude spectrum.

4. The blade status monitoring method according to claim 3, characterized in that: Constructing a memory matrix according to the reference frequency domain covariance and the reference operation monitoring data includes: Creating a benchmark data set using the benchmark frequency domain covariance and the benchmark operation monitoring data; Downsampling the benchmark data set to obtain downsampled benchmark frequency domain covariance and downsampled benchmark operation monitoring data; The downsampled benchmark frequency domain covariance and the downsampled benchmark operation monitoring data are combined to form the memory matrix.

5. The blade status monitoring method according to claim 4, characterized in that: The calculation of the baseline similarity threshold between any two sections includes: Eliminating the downsampled benchmark frequency domain covariance and the downsampled benchmark operation monitoring data from the benchmark data set to obtain a training data set; Select the data corresponding to any time T from the training data set as the observation vector ; Based on the observation vector Calculate the estimated vector ; Calculate the observation vector corresponding to any time T and the estimated vector Similarity between ; Repeat the above steps to calculate the similarity corresponding to each moment in the training data set to construct a similarity set , and based on the similarity set Get the baseline similarity threshold.

6. The blade status monitoring method according to claim 5, characterized in that: The observation vector based on Calculate the estimated vector Follow the following formula: , Wherein, D is the memory matrix.

7. The blade status monitoring method according to claim 5, characterized in that: The observation vector corresponding to any time T is calculated according to the following formula and the estimated vector Similarity between : , in, ; H is the total number of monitoring items of the benchmark operation monitoring data of the wind power generation equipment.

8. The blade status monitoring method according to claim 7, characterized in that: The calculating of the actual similarity value between any two sections according to the actual frequency domain covariance, the actual operation monitoring data and the memory matrix comprises: Construct the actual observation vector based on the actual frequency domain covariance and the actual operation monitoring data ; According to the actual observation vector and the memory matrix to calculate the actual estimated vector between any sections ; According to the actual observation vector and the actual estimated vector Calculate the actual similarity value between any two cross sections.

9. The blade status monitoring method according to claim 8, characterized in that: The reference similarity threshold includes a damage alarm threshold: If the actual similarity value is less than the damage alarm threshold, it indicates that the blade is damaged.

10. The blade status monitoring method according to claim 9, characterized in that: The damage alarm threshold is calculated by the following formula: , in, is the damage alarm threshold, is the alarm threshold coefficient; and The similarity sets are The mean and standard deviation of .

Images