Blade damage monitoring method and system of wind generating set, storage medium and program product

Abstract

The invention provides a blade damage monitoring method and system for a wind generating set, a storage medium and a program product, and the blade damage monitoring method comprises the steps: obtaining a voiceprint signal collected by a voiceprint sensor on a blade in the rotation process of the blade of the wind generating set; extracting voiceprint features for the voiceprint signal; according to the rotation period of the blade, detecting the characteristic value change of the voiceprint characteristic to obtain a detection result; and obtaining a blade damage monitoring result according to the detection result. The problem that the blade damage cannot be accurately monitored possibly due to noise of parts such as a transmission chain is solved, the influence of noise or vibration generated by the parts such as the transmission chain in a unit on the blade damage monitoring result can be reduced as much as possible, and the blade damage monitoring accuracy is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of wind power generation, and more specifically, to a blade damage monitoring method, system, storage medium and program product for a wind generator set. Background Art

[0002] With the development of wind power generation technology, soundprint monitoring technology is increasingly used in blade damage monitoring solutions.

[0003] Acoustic print monitoring is a non-destructive testing method for blade damage monitoring. When the outside of a wind turbine blade is damaged, it will cause changes in the vortex field distribution and aerodynamic noise characteristics, stimulating distinguishable acoustic print features; when the internal structure of the blade is damaged, the blade acoustic print features will also change along with the energy release of the elastic stress wave. Therefore, by collecting the acoustic print signals of the blades of a wind turbine and comparing and analyzing their acoustic print features, it is possible to find out whether the blades show signs of damage.

[0004] The use of soundprint monitoring technology has many advantages, such as all-weather monitoring, flexible installation of soundprint sensors, and convenient maintenance. However, since the soundprint signal is easily affected by the impact noise transmitted by the unit's transmission chain and other components, the presence of noise from the transmission chain and other components may lead to inaccurate monitoring results. Summary of the invention

[0005] In view of the problem in the related art that noise from components such as a transmission chain may lead to an inability to accurately monitor blade damage, the present disclosure provides a blade damage monitoring method, system, storage medium and program product for a wind turbine generator set.

[0006] A first aspect of the present disclosure provides a blade damage monitoring method for a wind turbine generator set, wherein a soundprint sensor is configured on the blade of the wind turbine generator set, wherein the blade damage monitoring method comprises: during the rotation of the blade of the wind turbine generator set, acquiring a soundprint signal collected by the soundprint sensor on the blade; extracting a soundprint feature for the soundprint signal, wherein a characteristic value of the soundprint feature represents an energy level of the soundprint signal; detecting a change in the characteristic value of the soundprint feature according to the rotation period of the blade to obtain a detection result, wherein the detection result represents a correlation between the change in the characteristic value of the soundprint feature and the rotation period of the blade; and obtaining a result of blade damage monitoring according to the detection result.

[0007] Optionally, the change in the characteristic value of the voiceprint feature is detected in the following manner: determining a characteristic value sequence of the voiceprint feature, wherein each characteristic value in the characteristic value sequence corresponds to a time series interval; determining a plurality of peak intervals formed by each characteristic value in the characteristic value sequence, wherein each peak interval contains a characteristic value greater than a peak threshold; and detecting the change in the characteristic value between each peak interval according to the rotation period of the blade to obtain the detection result.

[0008] Optionally, the changes in characteristic values ​​between each peak interval are detected in the following manner: determining a representative value of the characteristic value in each peak interval; obtaining a target time series based on the time value corresponding to the representative value of each peak interval; determining the difference between every two adjacent time values ​​in the target time series to obtain multiple time difference values; and obtaining the detection result by comparing the duration of the rotation cycle of the blade with each time difference value.

[0009] Optionally, the detection result is obtained in the following manner: in response to the difference between each of the multiple time difference values ​​and the duration of the rotation cycle of the blade being less than a first preset value, a detection result is obtained that shows a correlation between the change in the characteristic value characterizing the voice print feature and the rotation cycle of the blade; in response to the difference between at least one of the multiple time difference values ​​and the duration of the rotation cycle of the blade being greater than or equal to the first preset value, the at least one time difference value is compared with the durations of the multiple rotation cycles to obtain the detection result.

[0010] Optionally, the at least one time difference is compared with the duration of the multiple rotation cycles to obtain the detection result, including: in response to the fact that the difference between the value of each time difference in the at least one time difference divided into any number of parts and the duration of the rotation cycle is less than a second preset value, a detection result is obtained that there is a correlation between the change in the characteristic value of the voice print feature and the rotation cycle of the blade; in response to the presence of a target time difference in the at least one time difference, a detection result is obtained that there is no correlation between the change in the characteristic value of the voice print feature and the rotation cycle of the blade, wherein the difference between the value of the target time difference divided into any number of parts and the duration of the rotation cycle is greater than or equal to the second preset value.

[0011] Optionally, the peak interval is determined in the following manner: based on the eigenvalue sequence, a plurality of first eigenvalues ​​and a plurality of second eigenvalues ​​are determined, wherein the first eigenvalue is greater than the peak threshold and a previous eigenvalue of the first eigenvalue is less than or equal to the peak threshold, the second eigenvalue is less than or equal to the peak threshold and a previous eigenvalue of the second eigenvalue is greater than the peak threshold; the starting point of the peak interval is determined from the plurality of first eigenvalues, and the end point of the peak interval is determined from the plurality of second eigenvalues, so as to determine the peak interval.

[0012] Optionally, the multiple first eigenvalues ​​form a first sequence in time sequence, and the multiple second eigenvalues ​​form a second sequence in time sequence, wherein the peak interval is also determined in the following manner: for the target eigenvalue in the first sequence, the following operations are performed in sequence: based on the moment corresponding to the current target eigenvalue and the duration of the rotation cycle of the blade, a target time interval is determined, wherein the length of the target time interval is equal to the duration of the rotation cycle of the blade, and the target time interval contains the moment corresponding to the current target eigenvalue; in the first sequence, a first eigenvalue that is within the target time interval and closest to the interval starting point of the target time interval is searched to obtain a first target value; in the second sequence, a second eigenvalue that is within the target time interval and closest to the interval end point of the target time interval is searched to obtain a second target value; the first target value is used as the interval starting point of the current peak interval, the second target value is used as the interval end point of the current peak interval, and the first eigenvalue after the first target value in the first sequence is used as the next target eigenvalue, wherein the initial target eigenvalue is the first first eigenvalue in the first sequence.

[0013] Optionally, the result of blade damage monitoring is obtained based on the detection result, including: in response to the detection result characterizing that there is a correlation between the change in the characteristic value of the soundprint feature and the rotation period of the blade, and the number of characteristic values ​​of the characteristic value of the soundprint feature that is greater than a preset threshold is greater than or equal to a preset number, determining that the blade is at risk of damage.

[0014] Optionally, the voiceprint feature includes one of the following items: a statistical value of the amplitude of each frequency band interval in a plurality of frequency band intervals in the frequency spectrum of the voiceprint signal; a peak-to-peak value of the voiceprint signal; a sound pressure correlation value of the voiceprint signal; a kurtosis of the voiceprint signal; or an effective value of the voiceprint signal.

[0015] A second aspect of the present disclosure provides a blade damage monitoring system for a wind turbine generator set, the blade damage monitoring system comprising a monitoring platform and a soundprint sensor configured on the blades of the wind turbine generator set, the monitoring platform comprising: a processor; a memory for storing instructions executable by the processor, wherein the processor executable instructions, when executed by the processor, prompt the processor to execute the blade damage monitoring method for the wind turbine generator set according to the embodiment of the present disclosure.

[0016] Optionally, the wind turbine generator set also includes a main control system, and the monitoring platform sends the execution result of the processor to the main control system, so that the main control system performs a control action corresponding to the execution result, and the control action includes controlling the wind turbine generator set to retract blades and / or shut down.

[0017] A third aspect of the present disclosure provides a computer-readable storage medium. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the blade damage monitoring method of a wind turbine generator set according to an embodiment of the present disclosure.

[0018] A fourth aspect of the present disclosure provides a computer program product, comprising computer executable instructions, which, when executed by at least one processor, implement the blade damage monitoring method for a wind turbine generator set according to an embodiment of the present disclosure.

[0019] According to the blade damage monitoring scheme of the wind turbine generator set disclosed in the present invention, the soundprint signal can be obtained during the rotation of the blade, and the soundprint feature can be extracted to detect the characteristic value change of the soundprint feature according to the rotation period of the blade. The detection result can characterize the correlation between the characteristic value change of the soundprint feature and the rotation period of the blade, so as to obtain the monitoring result of the blade damage. In the case that the blade is damaged, the soundprint feature generated by the damage can be correlated with the rotation period of the blade, while the transmission chain noise has no necessary relationship with the rotation period of the blade. Therefore, by analyzing the soundprint feature in the above manner, it is possible to observe whether the characteristic value change of the soundprint feature is correlated with the periodicity of the blade rotation, so as to minimize the influence of the noise or vibration generated by the transmission chain and other components in the unit on the blade damage monitoring result, thereby improving the accuracy of blade damage monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 Schematic diagram showing the configuration position of a soundprint sensor in a blade damage monitoring method of a wind turbine generator system according to an exemplary embodiment of the present disclosure.

[0021] Figure 2is a schematic flow chart showing a method for monitoring blade damage of a wind turbine generator system according to an exemplary embodiment of the present disclosure.

[0022] Figure 3 is a schematic flowchart showing an example of extracting soundprint features in a method for monitoring blade damage of a wind turbine generator system according to an exemplary embodiment of the present disclosure.

[0023] Figure 4 1 is a schematic flowchart showing an example of the step of detecting the change of the characteristic value of the sound print feature in the blade damage monitoring method of the wind turbine generator according to the exemplary embodiment of the present disclosure.

[0024] Figure 5 is a schematic flow chart showing the steps of determining a peak interval in a method for monitoring blade damage of a wind turbine according to an exemplary embodiment of the present disclosure.

[0025] Figure 6 is a schematic flow chart showing operations performed on a target characteristic value in a blade damage monitoring method of a wind turbine generator system according to an exemplary embodiment of the present disclosure.

[0026] Figure 7 is a schematic flowchart showing another example of the step of detecting a change in a characteristic value of a sound print feature in a blade damage monitoring method of a wind turbine generator set according to an exemplary embodiment of the present disclosure.

[0027] Figure 8 is a schematic flow chart showing an application example of detecting changes in characteristic values ​​of sound print features in a blade damage monitoring method of a wind turbine generator set according to an exemplary embodiment of the present disclosure.

[0028] Fig. 9 is a schematic diagram showing the effect of periodic extraction of soundprint features in a blade damage monitoring method for a wind turbine generator system according to an exemplary embodiment of the present disclosure.

[0029] Fig.10 is a schematic diagram showing an architecture of a method for monitoring blade damage of a wind turbine generator system according to an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION

[0030] The following specific embodiments are provided to help the reader obtain a comprehensive understanding of the methods, devices and / or systems described herein. However, after understanding the disclosure of the present application, various changes, modifications and equivalents of the methods, devices and / or systems described herein will be clear. For example, the order of operations described herein is only an example and is not limited to those orders set forth herein, but can be changed as will be clear after understanding the disclosure of the present application, except for operations that must occur in a specific order. In addition, for greater clarity and simplicity, the description of features known in the art may be omitted.

[0031] The features described herein can be implemented in different forms and should not be construed as being limited to the examples described herein. Rather, the examples described herein have been provided to illustrate only some of the many possible ways to implement the methods, devices, and / or systems described herein, which will be clear after understanding the disclosure of the present application.

[0032] As used herein, the term "and / or" includes any one of the associated listed items and any combination of any two or more.

[0033] Although terms such as "first", "second", and "third" may be used herein to describe various members, components, regions, layers, or portions, these members, components, regions, layers, or portions should not be limited by these terms. Instead, these terms are only used to distinguish one member, component, region, layer, or portion from another member, component, region, layer, or portion. Therefore, without departing from the teachings of the examples described herein, the first member, first component, first region, first layer, or first portion referred to in the examples may also be referred to as the second member, second component, second region, second layer, or second portion.

[0034] In the specification, when an element (such as a layer, a region, or a substrate) is described as being “on”, “connected to”, or “coupled to” another element, the element may be directly “on”, “connected to”, or “coupled to” another element, or one or more other elements may be present therebetween. Conversely, when an element is described as being “directly on”, “directly connected to”, or “directly coupled to” another element, there may be no other elements present therebetween.

[0035] The terms used herein are only used to describe various examples and are not intended to limit the disclosure. Unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. The terms "comprise", "include" and "have" indicate the presence of the described features, quantities, operations, components, elements and / or combinations thereof, but do not exclude the presence or addition of one or more other features, quantities, operations, components, elements and / or combinations thereof.

[0036] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by a person of ordinary skill in the art to which the present disclosure belongs after understanding the present disclosure. Unless explicitly defined as such herein, terms (such as those defined in a general dictionary) should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and the present disclosure, and should not be interpreted in an idealized or overly formal manner.

[0037] Furthermore, in the description of examples, when it is considered that a detailed description of a well-known related structure or function would cause vague interpretation of the present disclosure, such a detailed description will be omitted.

[0038] As mentioned above, since the soundprint signal is easily affected by the impact noise transmitted by the transmission chain and other components of the unit, the presence of noise from the transmission chain and other components may lead to inaccurate monitoring results. In particular, judging the degree of blade failure by judging whether the characteristics at a single moment exceed the preset threshold is prone to false alarms.

[0039] In view of such problems, the present disclosure provides a blade damage monitoring method, a blade damage monitoring system, a computer-readable storage medium and a computer program product for a wind turbine generator set to solve or at least alleviate the problem that blade damage cannot be accurately monitored in the related art.

[0040] According to a first aspect of an exemplary embodiment of the present disclosure, a method for monitoring blade damage of a wind turbine is provided, wherein a soundprint sensor may be configured on the blade of the wind turbine.

[0041] Figure 1 The figure shows an example configuration position of a soundprint sensor on a blade according to an exemplary embodiment of the present disclosure. A wind turbine generator set may include, for example, three blades, and a soundprint sensor may be configured at the root of each blade. The installation position may be as follows: Figure 1 Here, the sampling frequency of the voiceprint sensor may be, for example, 44.1 KHz. As an example, the voiceprint sensor may be a sensor arranged in an existing health monitoring system of a wind turbine generator set.

[0042] Based on the soundprint sensor configured as described above, the blade damage monitoring method according to an embodiment of the present disclosure may include the following steps:

[0043] like Figure 2 As shown, in step S210, during the rotation of the blades of the wind turbine generator set, a soundprint signal collected by a soundprint sensor on the blades can be obtained.

[0044] Specifically, during the operation of the wind turbine generator set, for example, when the wind turbine generator set is in power generation operation, the soundprint signal can be collected by the soundprint sensor disposed on each blade.

[0045] As an example, after the voiceprint signal is acquired, the voiceprint signal may be preprocessed to obtain a processed signal for subsequent feature analysis.

[0046] In one example, preprocessing of the voiceprint signal may include pre-emphasis processing.

[0047] Specifically, when monitoring the blades of wind turbines on site, external factors may have a significant impact on the collected soundprint signals. Air, as the medium for sound signal propagation, consumes energy while propagating the sound signal energy. The soundprint energy is attenuated when propagating in the blade cavity. The higher the frequency of the signal, the more serious the loss of soundprint energy by the medium, which greatly weakens the energy of the high-frequency part of the collected soundprint signal, bringing a greater impact on signal processing.

[0048] Pre-emphasis technology is a commonly used sound signal processing method, which can be used to compensate for the characteristic of weak energy in the high-frequency part of the voiceprint signal and enhance the high-frequency part, thereby making the entire spectrum more balanced.

[0049] For example, assuming that the nth sampling point of the input signal is x[n], the pre-emphasis processing can be expressed by the following equation (1):

[0050] y[n]=x[n]-αx[n-1] (1)

[0051] Wherein, y[n] represents the output signal of the nth sampling point, α is a preset coefficient, and the value range of α can be [0.9, 1], for example, α=0.97.

[0052] Alternatively or additionally, in another example, the preprocessing of the voiceprint signal may include high-pass filtering.

[0053] Specifically, the energy of the original collected voiceprint signal or the voiceprint signal after pre-emphasis processing in the local frequency band (for example, the 0-800Hz frequency band) is high and the frequency components are complex, and the periodic energy concentration in the signal is not obvious. Therefore, in order to further improve the effect of subsequent processing, the voiceprint signal can be high-pass filtered.

[0054] For example, a Butterworth high-pass filter may be used, and the cut-off frequency may be selected as 2000 Hz, for example, to perform high-pass filtering on the voiceprint signal, and the voiceprint signal after filtering may only retain the high-frequency components in the signal.

[0055] Although it is described above that the collected soundprint signals can be preprocessed by pre-emphasis technology and / or high-pass filtering to further improve the monitoring effect, the embodiments of the present disclosure are not limited to this, and the above-mentioned preprocessing may not be used. Without the above-mentioned preprocessing, soundprint feature extraction and detection can be achieved, and accurate monitoring of blade damage can be achieved.

[0056] In step S220, voiceprint features may be extracted from the voiceprint signal, wherein the characteristic value of the voiceprint feature may represent the energy level of the voiceprint signal.

[0057] In this step, the voiceprint features that can characterize the energy level of the voiceprint signal can be extracted to analyze whether there is a periodicity in the energy change related to the blade rotation cycle.

[0058] Specifically, when a blade is damaged, a periodic energy concentration phenomenon usually occurs in the time-frequency diagram of the blade soundprint signal. This energy concentration phenomenon is more obvious at high frequencies (for example, above 3KHz). In the early stages of blade damage, the periodic energy components at high frequencies are relatively weak; when the damage expands, the energy of the periodic energy concentration components in the signal time-frequency diagram increases rapidly and can even cover the entire frequency band. Therefore, by identifying the periodic energy concentration components in the early stages of blade failure and continuously monitoring their development, the correlation between energy changes and the blade rotation period can be detected, which can more accurately monitor blade damage.

[0059] As an example, the voiceprint features may include one or more of the following items: the statistical value of the amplitude of each frequency band interval in multiple frequency band intervals in the frequency spectrum of the voiceprint signal; the peak-to-peak value of the voiceprint signal; the sound pressure related quantity of the voiceprint signal; the kurtosis of the voiceprint signal; and the effective value of the voiceprint signal.

[0060] The following will refer to Figure 3 The process of extracting voiceprint features is described by taking the statistical values ​​of amplitudes of multiple frequency band intervals as an example.

[0061] As an example, when the unit is generating electricity normally, it usually takes 5 to 12 seconds for the blade to rotate one circle. In order to improve the accuracy of automatically judging the periodic energy concentration phenomenon in the signal, the duration of a selected signal can cover more than 2 blade rotation cycles as much as possible. Therefore, here, the voiceprint data with a duration of about 28 seconds is initially selected as a signal to be analyzed.

[0062] like Figure 3As shown, in step S301, calculation scheduling can be performed every predetermined calculation cycle to obtain the collected soundprint signal. Here, the calculation cycle can be, for example, 60 seconds, that is, data can be input once every 60 seconds, and the data input duration in each scheduling can be 28 seconds. In step S302, it can be determined whether the current state is a power generation state. The method can be executed in the power generation state of the unit, so that blade damage can be more accurately identified in the regular and continuous rotation of the blade.

[0063] In response to the current non-power generation condition, the process can return to step S301 to continue acquiring input data. In response to the current power generation condition, in step S303, the data can be framed and windowed. Specifically, a voiceprint signal data with a time length of 28 seconds can be framed and windowed. The parameters are frame length (or window length) nframe = 200ms, frame shift (or window movement distance) n_hop = 100ms, and the selected window function can be, for example, a Hamming window. Although an example in which the frame shift is less than the frame length is described here, that is, the time window of each windowing overlaps with the time window of the previous windowing, it is not limited to this, and the frame shift can also be greater than or equal to the frame length.

[0064] In step S304, the voiceprint feature can be calculated frame by frame. For example, taking the voiceprint feature as the statistical value of the amplitude of each frequency band interval (also referred to as "frequency band") in multiple frequency band intervals as an example, the statistical value of the amplitude of different frequency band intervals in the voiceprint signal spectrum can be determined. In steps S305 and S306, the change of the characteristic value can be detected to determine whether the blade is damaged, which will be described in detail below.

[0065] As an example, the statistic of the amplitude may be the sum of the amplitudes, but it is not limited thereto, and other statistical values ​​may also be selected, such as the mean, median, mode, etc. In addition, the embodiments of the present disclosure do not particularly limit the number of frequency intervals and the way of dividing the intervals. For example, the six voiceprint features shown in Table 1 below may be processed, that is, six frequency intervals may be selected: 2-3kHz, 3-6kHz, 6-9kHz, 9-12kHz, 12-16kHz, and 16-18kHz, and the statistical values ​​of the amplitudes in each interval are respectively counted. For example, as shown in Table 1, the sum of the amplitudes in each interval is counted.

[0066] Table 1

[0067]

[0068]

[0069] The calculation methods of statistical values ​​of different frequency intervals may be similar. For example, taking the six voiceprint features in Table 1 above as an example, the statistical methods of features Sum_amp1, Sum_amp2, Sum_amp3, Sum_amp4, Sum_amp5 and Sum_amp6 may be similar.

[0070] Specifically, for the voiceprint signal after framing in any frequency interval, Fourier transform (FFT) can be performed starting from the first frame data to obtain the amplitude spectrum of the frame data. The amplitudes of the frequency band range in the amplitude spectrum within the frequency interval are added to obtain the sum of the amplitude characteristics of the specific frequency band of this frame data. The characteristic values ​​of the voiceprint features are calculated with a frame length of 200ms and a frame shift of 100ms to obtain the characteristic values ​​of the voiceprint features, and the obtained characteristic values ​​can be spliced ​​into the voiceprint feature Sum_amp of the voiceprint signal in this interval in the order of the frame numbers. The data length of the voiceprint feature Sum_amp is the total number of frames, and the corresponding time length is consistent with the original signal before framing (for example, 28 seconds). Therefore, a time series time_list corresponding to the voiceprint feature can be generated, and the time series can contain the time corresponding to each voiceprint feature value.

[0071] Here, considering that there may be more than one periodic impact of the blade soundprint within a certain time period, the energy distribution of different types of periodic impacts in the blade soundprint may be different. Therefore, in this example, by calculating the amplitude energy and characteristics of the blade soundprint data in frequency bands, multiple periodic impact phenomena occurring within a certain time period can be identified.

[0072] In addition to using the amplitude statistics of the above-mentioned spectrum as voiceprint features, in the embodiments of the present disclosure, the voiceprint features may also include time domain features, such as the peak-to-peak value of the voiceprint signal, the sound pressure related amount of the voiceprint signal, the kurtosis of the voiceprint signal and the effective value of the voiceprint signal. For these features, the data can be framed and windowed in a similar manner to the above to obtain the characteristic values ​​corresponding to each frame, and its time series can correspond to the time series time_list of the above-mentioned frequency domain voiceprint features. Therefore, in the embodiments of the present disclosure, the voiceprint features may include time domain features, frequency domain features or a combination of the two.

[0073] Although some examples of relevant voiceprint features of voiceprint signal processing are given above, the embodiments of the present disclosure are not limited to the above examples. No matter how the features are combined, which feature quantification method is adopted, and the different feature calculation times are different, as long as the phenomenon of periodic impact in the blade soundprint (such as the spectrogram) is identified and used as a judgment condition for blade damage, it can be applied to the embodiments of the present disclosure.

[0074] In step S230, the characteristic value change of the voiceprint feature can be detected according to the rotation period of the blade to obtain a detection result.

[0075] Here, the detection result can characterize the correlation between the characteristic value change of the voiceprint feature and the rotation period of the blade. As an example, taking the above Table 1 as an example, the six voiceprint features can be periodically detected respectively. When at least one feature is correlated with the blade rotation period, it can be determined that there is periodic energy concentration in the voiceprint signal. For example, a feature of the voiceprint signal (such as the above feature Sum_amp1) and its corresponding time series (such as the time series time_list1) and the blade rotation period T can be used as inputs for periodic detection.

[0076] As an example, the change in the characteristic value of the voiceprint feature can be detected in the following way:

[0077] like Figure 4 As shown, in step S410, a feature value sequence of the voiceprint feature may be determined, wherein each feature value in the feature value sequence may correspond to a time series interval.

[0078] As described above, the characteristic value of the voiceprint feature can be obtained by frame windowing, and a characteristic value can be obtained each time the window is slid, and multiple characteristic values ​​in the characteristic value sequence can correspond to multiple time series intervals respectively. The time series interval mentioned here can be the time window where the above-mentioned windowed window is located. The characteristic value sequence can include, for example, the characteristic value of the above-mentioned feature Sum_thre1 counted in each window.

[0079] In step S420, a plurality of peak intervals formed by the eigenvalues ​​in the eigenvalue sequence may be determined, wherein each peak interval may contain eigenvalues ​​greater than a peak threshold.

[0080] In this step, multiple peak intervals can be obtained by selecting eigenvalues ​​greater than the peak threshold, so that each peak interval has a larger eigenvalue, or an eigenvalue with more concentrated energy, which is beneficial for subsequent periodic detection of the eigenvalue.

[0081] As an example, the peak threshold may be preset, such as an empirical value.

[0082] As another example, the peak threshold may also be determined based on the feature value of the current voiceprint feature, for example, the peak threshold may be determined based on the maximum value in the current feature value sequence, for example, the peak threshold Sum_thre1 = k × max(Sum_amp1), where max(Sum_amp1) represents the maximum value in the feature sequence (e.g., Sum_amp1), and k represents a preset coefficient, for example, 0.5. However, the embodiments of the present disclosure are not limited thereto, and the peak threshold may also be determined based on the minimum value, average value, median, mode, etc. in the current feature value sequence.

[0083] When determining the peak threshold, in one example, the interval between two adjacent characteristic values ​​greater than the peak threshold can be taken as the peak interval. In another example, the embodiment of the present disclosure can also merge the peak intervals to ensure that there is only one peak interval within the duration of a single blade rotation cycle.

[0084] Specifically, the peak interval can be determined in the following ways:

[0085] like Figure 5 As shown, in step S510, a plurality of first eigenvalues ​​and a plurality of second eigenvalues ​​may be determined based on the eigenvalue sequence.

[0086] Here, the first eigenvalue may be greater than the peak threshold and a previous eigenvalue of the first eigenvalue is less than or equal to the peak threshold, and the second eigenvalue may be less than or equal to the peak threshold and a previous eigenvalue of the second eigenvalue is greater than the peak threshold.

[0087] Specifically, the first eigenvalue can be a eigenvalue in the eigenvalue sequence that is itself greater than the peak threshold but its previous value is less than the peak threshold. Such a eigenvalue may be the starting point of the peak interval where energy is concentrated; the second eigenvalue can be a eigenvalue in the eigenvalue sequence that is itself less than the peak threshold but its previous value is greater than the peak threshold. Such a eigenvalue may be the end point of the peak interval where energy is concentrated.

[0088] As an example, the peak threshold can be subtracted from each value in the feature value sequence to obtain the first feature sequence sum_temp. According to the value in the first feature sequence sum_temp, if the value is greater than 0, it is set to 1, and if the value is less than or equal to 0, it is set to zero, so as to convert the first feature sequence sum_temp into a second feature sequence containing only 0 and 1.

[0089] Based on the second feature sequence, the first mutation point position index (index_st) that changes from 0 to 1 and the second mutation point position index (index_end) that changes from 1 to 0 in the second feature sequence can be searched. The eigenvalue corresponding to each position in the first mutation point position index index_st can be used as the first eigenvalue, which may be the starting point of a peak interval; the eigenvalue corresponding to each position in the second mutation point position index index_end can be used as the second eigenvalue, which may be the end point of a peak interval. In this way, the first eigenvalue and the second eigenvalue can be determined from the eigenvalue sequence. However, the embodiments of the present disclosure are not limited to this, and each eigenvalue in the eigenvalue sequence can also be searched one by one according to the first eigenvalue being greater than the peak threshold and the previous eigenvalue of the first eigenvalue being less than or equal to the peak threshold, the second eigenvalue being less than or equal to the peak threshold and the previous eigenvalue of the second eigenvalue being greater than the peak threshold, to determine the first eigenvalue and the second eigenvalue.

[0090] In step S520, a starting point of the peak interval may be determined from the plurality of first eigenvalues, and an end point of the peak interval may be determined from the plurality of second eigenvalues, so as to determine the peak interval.

[0091] Here, the peak interval can be determined based on the interval formed by the first eigenvalue and the second eigenvalue. Since there are multiple first eigenvalues ​​and multiple second eigenvalues, adjacent first eigenvalues ​​and second eigenvalues ​​can form a peak interval, thereby obtaining multiple peak intervals. In this case, when determining the peak interval, it is possible to make only one peak interval exist within the duration of a single rotation cycle of the blade. In this way, the rotation cycle of the blade can have the same frequency of occurrence as the peak interval, so that it is possible to better analyze whether there is a correlation between the cycles of the two.

[0092] It should be noted that the duration of a single rotation cycle mentioned here refers to the same as the duration of the rotation cycle in time, but does not mean that there is a complete peak interval in each rotation cycle.

[0093] As an example, a plurality of first eigenvalues ​​may form a first sequence in time sequence, and a plurality of second eigenvalues ​​may form a second sequence in time sequence, wherein the peak interval may also be determined in the following manner: for the target eigenvalue in the first sequence, the following operations are performed in sequence:

[0094] like Figure 6 As shown, in step S610, the target time interval may be determined based on the moment corresponding to the current target characteristic value and the duration of the rotation cycle of the blade.

[0095] Here, the length of the target time interval may be equal to the duration of the rotation period of the blade, and the target time interval may include the time corresponding to the current target characteristic value.

[0096] For example, the time corresponding to the current target characteristic value can be used as the center time of the target time interval, half of the duration of the blade rotation cycle can be taken forward, and half of the duration of the blade rotation cycle can be taken backward, and the obtained time interval can be used as the target time interval. However, it is not limited to this, and a times of the duration of the blade rotation cycle can be taken forward, and (1-a) times of the duration of the blade rotation cycle can be taken backward, where the value of a is in the range of [0,1].

[0097] In step S620, a first characteristic value that is within the target time interval and closest to the starting point of the target time interval may be searched in the first sequence to obtain a first target value.

[0098] In step S630, a second characteristic value that is within the target time interval and closest to the end point of the target time interval may be searched in the second sequence to obtain a second target value.

[0099] In the case of determining the target time interval, since the duration of a single target time interval can be equal to the duration of the rotation cycle of the blade, only one peak interval can be determined in the single target time interval. In this case, the first characteristic value that is within the target time interval and closest to the starting point of the target time interval can be taken from the first sequence, and the second characteristic value that is within the target time interval and closest to the end point of the target time interval can be taken from the second sequence.

[0100] In step S640, the first target value may be used as the interval starting point of the current peak interval, the second target value may be used as the interval ending point of the current peak interval, and the first characteristic value after the first target value in the first sequence may be used as the next target characteristic value.

[0101] Here, the initial target eigenvalue may be the first first eigenvalue in the first sequence.

[0102] In the above manner, each first eigenvalue in the first sequence can be searched in order to determine each peak interval. Although the above describes an example of sequentially searching for each first eigenvalue in the first sequence to determine each peak interval, the embodiments of the present disclosure are not limited thereto, and each second eigenvalue in the second sequence can also be searched in order to determine each peak interval, and the process is similar to the above process and will not be repeated here.

[0103] In addition, the process of determining the peak interval is not limited to the above example. According to the embodiment of the present disclosure, considering that the characteristic value of the voiceprint feature may continuously appear multiple local peaks exceeding the threshold in a short period of time, for example, according to the rotation period T of the blade, the peak intervals can be selectively merged, and the values ​​of index_st and index_end can be updated to ensure that only one characteristic peak is extracted within the duration of a blade rotation period in subsequent processing. For example, the difference between the time t1 corresponding to the end position of the previous peak interval and the time t2 of the start position of the adjacent peak interval can be determined. If the difference is less than 0.5 times the blade rotation frequency, the two intervals are directly merged. According to the updated index value, multiple peak intervals are obtained in the characteristic value sequence.

[0104] Return to reference Figure 4 In step S430, the characteristic value changes between each peak interval can be detected according to the rotation period of the blade to obtain a detection result.

[0105] In the above process, multiple peak intervals can be determined, and according to the changes in the characteristic values ​​of each peak interval, the correlation with the rotation period of the blade can be detected. Thus, the overall situation of the soundprint signal can be characterized by the changes in the characteristic values ​​between the peak intervals, which simplifies the calculation amount of characteristic period monitoring and improves the recognition speed of blade damage.

[0106] In this step, as an example, the characteristic value change between intervals can be determined based on the representative value of each peak interval. For example, the characteristic value change between each peak interval can be detected in the following manner:

[0107] like Figure 7 As shown, in step S710, a representative value of the characteristic value in each peak interval can be determined.

[0108] As an example, the representative value of the peak interval may be a characteristic value at a preset position in the peak interval, such as a peak value in the peak interval, but it is not limited thereto. For example, the representative value may also be the median of the peak interval.

[0109] In step S720, a target time series may be obtained based on the time values ​​corresponding to the representative values ​​of each peak interval.

[0110] Here, the time values ​​corresponding to the representative values ​​can be obtained from the above time series, and the time values ​​corresponding to each representative value form the above target time series. In this way, a representative value sequence (e.g., peak sequence peak_amp1) and a representative value time series corresponding to the representative values ​​(e.g., peak time series peak_time1 corresponding to the peak sequence) can be obtained.

[0111] In step S730, the difference between every two adjacent time values ​​in the target time series may be determined to obtain a plurality of time difference values.

[0112] Two adjacent time values ​​in the target time series represent the time difference between the representative values ​​of two adjacent peak intervals, which can characterize the periodicity of the voiceprint feature. The time difference between the two representative values ​​can be regarded as the period of the voiceprint feature.

[0113] In step S740, the detection result may be obtained by comparing the duration of the rotation cycle of the blade with each time difference.

[0114] In the above manner, the time difference between the representative values ​​of adjacent peak intervals can be determined, and the time difference can be compared with the blade rotation period. When the time difference is close to the blade rotation period or is the same as the blade rotation period, it can be considered that the occurrence of characteristic values ​​with special energy distribution (such as peak values ​​of concentrated energy) is periodic, and its period matches the blade rotation period. In this case, it can be considered that there may be a risk of blade damage or further testing is needed. In this way, the influence of components such as the unit transmission chain on blade damage monitoring can be eliminated, thereby improving the accuracy of monitoring.

[0115] In step S740, the duration of the rotation cycle of the blade may be compared to see whether it matches or is close to the above time difference.

[0116] For example, the detection result can be obtained in the following manner: in response to the difference between each time difference value among multiple time difference values ​​and the duration of the rotation cycle of the blade being less than a first preset value, a detection result is obtained that shows a correlation between the change in the characteristic value characterizing the voice print feature and the rotation cycle of the blade; in response to the difference between at least one time difference value among multiple time difference values ​​and the duration of the rotation cycle of the blade being greater than or equal to the first preset value, the at least one time difference value is compared with the durations of multiple rotation cycles to obtain a detection result.

[0117] Here, the first preset value can be set according to actual needs, for example, it can be 0.4 seconds.

[0118] Specifically, the representative value time series (eg, peak_time1) corresponding to the representative value obtained in step S720 can be used to calculate the difference between adjacent elements in the representative value time series to obtain a difference array diff_list. Whether there is periodicity is determined by the distribution of each difference in the difference array diff_list.

[0119] For example, each difference in the difference array diff_list can be subtracted from the blade rotation period T to obtain the comparison difference diff_temp, and it is determined whether the absolute values ​​of the comparison differences diff_temp corresponding to each difference are all less than the first preset value. In response to the comparison differences being less than the first preset value, it is considered that the time series corresponding to the representative value has a periodicity that matches the blade rotation period; in response to at least one comparison difference being greater than or equal to the first preset value, in one example, it can be considered that the time series corresponding to the representative value does not have a periodicity that matches the blade rotation period.

[0120] In another example, when at least one comparison difference is greater than or equal to a first preset value, the case where a single time difference corresponds to multiple rotation cycles of the blade can also be excluded. For example, it can be further determined whether these time difference values ​​whose comparison difference is greater than or equal to the first preset value correspond to the duration of multiple rotation cycles to obtain a detection result.

[0121] Specifically, the detection result can be obtained in the following manner: in response to the fact that the difference between the value of each time difference value divided into any number of parts and the duration of the rotation cycle of the blade is less than the second preset value, a detection result is obtained that there is a correlation between the characteristic value change characterizing the voiceprint feature and the rotation cycle of the blade; in response to the presence of a target time difference value in the at least one time difference value, a detection result is obtained that there is no correlation between the characteristic value change characterizing the voiceprint feature and the rotation cycle of the blade, wherein the difference between the value of the target time difference value divided into any number of parts and the duration of the rotation cycle is greater than or equal to the second preset value. Here, the number of parts n can be a positive integer, for example, 2.

[0122] As an example, the difference diff in diff_list whose difference with the duration of the blade rotation cycle is greater than or equal to the first preset value can be divided by any positive integer. If the difference between these difference diffs and the duration of the blade rotation cycle after being divided by a certain positive integer is less than the second preset value, it is judged that the time series corresponding to the peak value has periodicity; if there is at least one difference diff among these difference diffs, no matter which positive integer it is divided by, the difference between it and the duration of the blade rotation cycle is greater than or equal to the second preset value, it is judged that the time series corresponding to the peak value does not have periodicity. Here, the second preset value can be equal to the first preset value, for example. In the above manner, when at least one comparison difference is greater than or equal to the first preset value, it can be checked whether there are time differences corresponding to multiple rotation cycles, so that the situation where a single time difference corresponds to multiple rotation cycles can be identified, avoiding underreporting, and further improving the accuracy of monitoring.

[0123] In addition, as an example, the method of an embodiment of the present disclosure may also include: in response to the number of elements in the difference array diff_list being greater than a preset number (for example, 2), executing the step of obtaining the detection result by comparing the duration of the rotation cycle of the blade with each time difference; in response to the number of elements in the difference array diff_list being less than or equal to the preset number, ending the blade damage monitoring process.

[0124] Although an example process for detecting changes in the eigenvalues ​​of voiceprint features is described above, the embodiments of the present disclosure are not limited thereto. For example, methods such as the autocorrelation coefficient method, random forest algorithm, long short-term memory neural network, STL (Seasonal and Trend Decomposition using Loess) decomposition, etc. may be used to determine the periodicity of the eigenvalues ​​of voiceprint features. For example, the periodic impact phenomenon in the blade voiceprint spectrogram is identified. When the periodicity is identified, its periodicity can be matched with the blade rotation period to determine the correlation between the changes in the eigenvalues ​​of the voiceprint features and the blade rotation period to obtain a detection result.

[0125] In step S240, the blade damage monitoring result may be obtained according to the detection result.

[0126] As an example, in response to the detection result that there is a correlation between the change in the characteristic value of the soundprint feature and the rotation period of the blade, it can be determined that the blade is at risk of damage; in response to the detection result that there is no correlation between the change in the characteristic value of the soundprint feature and the rotation period of the blade, it can be determined that the blade is not at risk of damage.

[0127] As another example, when the detection result indicates that there is a correlation between the change in the characteristic value of the voiceprint feature and the rotation period of the blade, it can be further determined whether all the characteristic values ​​of the voiceprint feature are greater than a preset threshold to further improve the accuracy of the monitoring result.

[0128] Specifically, in response to the detection result that the change in the characteristic value of the voiceprint feature characterizes that there is a correlation with the rotation period of the blade, and the number of characteristic values ​​of the voiceprint feature greater than a preset threshold is greater than or equal to the preset number, it can be determined that the blade is at risk of damage; in response to the detection result that the change in the characteristic value of the voiceprint feature characterizes that there is a correlation with the rotation period of the blade, and the number of characteristic values ​​of the voiceprint feature greater than the preset threshold is less than the preset number, it can be determined that the blade is not at risk of damage.

[0129] For example, Figure 3As shown, after calculating 6 voiceprint features frame by frame in step S304, in step S305, the periodicity of each voiceprint feature can be judged, and in step S306, in response to the correlation between the existence of at least one voiceprint feature and the blade rotation period, and among these voiceprint features with correlation, the number of feature values ​​greater than a preset threshold in the feature values ​​of at least one voiceprint feature is greater than or equal to a preset number, in step S307, it can be considered that the blade is damaged.

[0130] In addition, as an example, the preset number can be set according to actual needs, for example, it can be 2.

[0131] Here, the preset threshold can be set according to actual needs, and the preset thresholds corresponding to different voiceprint features can be different. Specifically, taking the frequency domain voiceprint features in Table 1 as an example, different voiceprint features have different frequency band energy distributions, and the thresholds peak_thre(i) of different features are also different.

[0132] In one example, when there is no blade failure soundprint data, a large amount of soundprint data collected from normal blades can be used to extract each soundprint feature and its corresponding peak feature sequence peak_amp(i), and the multiple (e.g., 2 times) of the minimum value in the obtained peak_amp(i) is used as the preliminary threshold peak_thre(i) of the corresponding soundprint feature. In another example, after collecting the soundprint data of the blade failure case, each soundprint feature and its corresponding representative value feature sequence peak_amp(i) are extracted, and the minimum value in the obtained representative value feature sequence peak_amp(i) is multiplied by a preset coefficient less than 1 (e.g., 0.25) as the threshold peak_thre(i).

[0133] Figure 8 A schematic flowchart of an application example of detecting changes in characteristic values ​​of sound print features in a blade damage monitoring method for a wind turbine generator set according to an exemplary embodiment of the present disclosure is shown.

[0134] like Figure 8 As shown, in step S801, a characteristic value sequence of the voiceprint feature can be extracted according to the voiceprint signal, and in this step, a time series and a blade rotation period corresponding to the characteristic value sequence can also be determined. In step S802, a peak threshold can be generated, for example, it can be determined for the maximum value in the characteristic value sequence. In step S803, a peak interval in which the characteristic value is greater than the peak threshold can be determined. In step S804, a maximum value in each peak interval and a corresponding time list can be determined.

[0135] In step S805, it can be determined based on the time list whether the time difference in the time list matches the rotation period of the blade. In response to the time difference matching the rotation period of the blade, in step S806, it can be determined that there is a correlation between the voice print feature and the rotation period of the blade. In response to the time difference not matching the rotation period of the blade, in step S807, it can be determined that there is no correlation between the voice print feature and the rotation period of the blade.

[0136] According to the embodiments of the present disclosure, by identifying whether there are periodic impacts in the soundprint characteristics of the blades, judging the degree of proximity of the period to the rotation period of the blades, and then using whether the amplitudes of the periodic characteristics all exceed the set threshold to preliminarily judge the damage of the blades, it is possible to better match the operating state of the blades and is not easily affected by the noise of the transmission chain.

[0137] Fig. 9 The figure shows the periodic effect of extracting the six soundprint features in Table 1 for the soundprint data of a blade without obvious damage, where the horizontal axis represents time (the unit can be, for example, seconds), and the vertical axis can represent the amplitude, such as the result of summing the frequency domain amplitude of the eigenvalue, and its unit can be determined based on the calculation of the eigenvalue. This is intended to show the effect of feature extraction, and there is no special restriction on the units of the horizontal and vertical axes. Fig. 9 In the figure, the hollow circles indicate the extracted representative value sequence peak_amp(i), and the data above each hollow circle represents the time corresponding to the point. From the trends of multiple voiceprint features, it can be found that there is a periodic feature enhancement of about 7.3 seconds. In this case, the period (7.3s) calculated by the algorithm is compared with the rotation period of the blade at this time. Since the difference between the two is less than the set first preset value (for example, 0.4s), it is judged that this signal has a periodic impact. However, since the peak sequence of each feature does not exceed the corresponding threshold peak_thre(i), it can only indicate that the voiceprint signal has periodicity, and no damage to the blade is detected.

[0138] According to the embodiments of the present disclosure, the soundprint data collected by the soundprint sensor of the blade is used to process the soundprint signal characteristics in frequency bands, and an algorithm is used to determine whether the characteristics are correlated with the rotation cycle of the blade. When the soundprint signal characteristics of the blade are correlated with the rotation cycle of the blade, a warning can be issued in time in the system and fault information can be saved; when the characteristic values ​​of the soundprint characteristics of the soundprint signal exceed the preset threshold, it may be necessary to shut down as soon as possible to avoid the rapid deterioration of the blade fault or even breakage.

[0139] In addition, according to the embodiments of the present disclosure, a judgment scheme for the periodic energy enhancement (or impact) of the blade sound pattern is proposed. By analyzing the relationship between the periodicity of the characteristics of each frequency band of the signal and the blade rotation period, the development trend of the characteristics of each frequency band can be continuously monitored from the early stage of blade failure. When the risk of blade damage is identified, a signal is transmitted to the main control system of the unit to realize the retraction and shutdown of the propeller to protect the blades, effectively reducing the risk of blade breakage caused by the expansion of blade damage.

[0140] According to a second aspect of an exemplary embodiment of the present disclosure, a blade damage monitoring system for a wind turbine generator set is provided. The wind turbine generator set includes a blade 10, such as Fig.10 As shown, the blade damage monitoring system may include a monitoring platform 1010 and a soundprint sensor 1020 configured on each blade. The monitoring platform 1010 may include: a processor; a memory for storing processor executable instructions, wherein the processor executable instructions, when executed by the processor, prompt the processor to execute the blade damage monitoring method of the wind turbine generator set according to the embodiment of the present disclosure.

[0141] As an example, the wind turbine generator set also includes a main control system 20, and the monitoring platform 1010 can send the execution result of the processor to the main control system 20, so that the main control system 20 performs a control action corresponding to the execution result, and the control action may include, for example, controlling the wind turbine generator set to retract the blades and / or shut down. For example, the monitoring platform 1010 may include a data collector for receiving signals from the soundprint sensor, and the data collector may send the data of the received signal to the processor. The main control system 20 can determine whether it is necessary to perform actions such as retracting the blades and shutting down according to the alarm or shutdown signal transmitted by the monitoring platform 1010, thereby realizing real-time damage protection for the blades.

[0142] Specifically, the monitoring platform 1010 can receive data from the soundprint sensor of each blade, and can receive the blade rotation period signal transmitted by the main control system 20, so that feature calculation and blade damage identification can be performed according to the monitoring method of the embodiment of the present disclosure.

[0143] The monitoring platform 1010 can also transmit the fault diagnosis results to the main control system. If the signal only shows periodic characteristics that match the blade rotation period, and the characteristic value does not exceed the preset threshold, the monitoring platform 1010 can transmit a warning signal to the wind turbine main control system 20; if the signal characteristics show periodic characteristics that match the blade rotation period, and the characteristic value exceeds the preset threshold, the monitoring platform 1010 can transmit a fault signal to the main control system 20.

[0144] In addition, the monitoring platform 1010 can also transmit the original voiceprint data, feature calculation results, blade damage judgment results and other information to the data storage server 30 of the wind farm where the wind turbine generator set is located. The data storage server 30 can be a data storage server for the entire wind farm, which uniformly stores all the voiceprint sensor data, feature calculation results, and blade damage status of all the units.

[0145] Specifically, the voiceprint sensor 1020 can provide access to the voiceprint data required to calculate all indicators; the monitoring platform 1010 can perform relevant logical judgments, including voiceprint feature processing, difference degree calculation, etc., and output corresponding diagnostic results, control instructions, etc. The main control system 20 can receive the monitoring results from the monitoring platform 1010, and can perform actions such as retracting and shutting down the blades according to the alarm or shutdown instructions corresponding to the monitoring results, thereby realizing real-time online protection of blade damage.

[0146] In addition, the wind farm where the wind turbine generator set is located can also be equipped with an equipment management system, which can uniformly manage information such as the status of the voiceprint sensor, the status of the monitoring platform, and the blade damage status of each unit in the entire wind farm.

[0147] Here, the monitoring platform 1010 may include an electronic device, which may include the processor and memory described above to execute the blade damage monitoring method of a wind turbine generator set.

[0148] As an example, the electronic device may be a PC, a tablet device, a personal digital assistant, a smart phone, or other device capable of executing the above instruction set. Here, the electronic device is not necessarily a single electronic device, but may also be any device or circuit capable of executing the above instruction (or instruction set) individually or in combination. The electronic device may also be part of an integrated control system or system manager, or may be configured as a portable electronic device interconnected with a local or remote (e.g., via wireless transmission) interface.

[0149] In electronic devices, the processor may include a central processing unit (CPU), a graphics processing unit (GPU), a programmable logic device, a dedicated processor system, a microcontroller or a microprocessor. As an example and not limitation, the processor may also include an analog processor, a digital processor, a microprocessor, a multi-core processor, a processor array, a network processor, etc.

[0150] The processor can execute instructions or codes stored in the memory, wherein the memory can also store data. Instructions and data can also be sent and received through the network via the network interface device, wherein the network interface device can adopt any known transmission protocol.

[0151] The memory may be integrated with the processor, for example, RAM or flash memory is arranged within an integrated circuit microprocessor or the like. In addition, the memory may include a separate device, such as an external disk drive, a storage array, or any other storage device that can be used by a database system. The memory and the processor may be operatively coupled, or may communicate with each other, such as through an I / O port, a network connection, etc., so that the processor can read files stored in the memory.

[0152] In addition, the electronic device may also include a video display (such as a liquid crystal display) and a user interaction interface (such as a keyboard, a mouse, a touch input device, etc.) All components of the electronic device may be connected to each other via a bus and / or a network.

[0153] According to a third aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided. Specifically, the blade damage monitoring method of a wind turbine generator set according to an embodiment of the present disclosure can be written as a computer program and stored on a computer-readable storage medium. When the instructions in the computer-readable storage medium are executed by at least one processor, the at least one processor is prompted to execute the blade damage monitoring method of the wind turbine generator set according to the exemplary embodiment of the present disclosure. Examples of computer-readable storage media include: read-only memory (ROM), random access programmable read-only memory (PROM), electrically erasable programmable read-only memory (EEPROM), random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), flash memory, non-volatile memory, CD-ROM, CD-R, CD+R, CD-RW, CD+RW, DVD-ROM, DVD-R, DVD+R, DVD-RW, DVD+RW, DVD-RAM, BD-ROM, BD-R, BD-R LTH, BD-RE, Blu-ray or optical disk storage, hard disk drive (HDD), solid state drive (SSD), card storage (such as, multimedia card, secure digital (SD) card or extreme digital (XD) card), magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid state disk and any other device, any other device is configured to store computer program and any associated data, data files and data structures in a non-transitory manner and provide the computer program and any associated data, data files and data structures to a processor or computer so that the processor or computer can execute the computer program. In one example, the computer program and any associated data, data files and data structures are distributed on a networked computer system so that the computer program and any associated data, data files and data structures are stored, accessed and executed in a distributed manner by one or more processors or computers.

[0154] According to a fourth aspect of an embodiment of the present disclosure, a computer program product is provided. The computer program product includes computer instructions. When the computer instructions are executed by a processor, the blade damage monitoring method of a wind turbine generator set according to an embodiment of the present disclosure can be implemented.

[0155] The specific implementation methods of the present disclosure have been described in detail above. Although some embodiments have been shown and described, those skilled in the art should understand that these embodiments may be modified and varied without departing from the principles and spirit of the present disclosure, the scope of which is defined by the claims and their equivalents. These modifications and variations should also be within the scope of protection of the claims of the present disclosure.

Claims

1. A method for monitoring blade damage of a wind turbine generator set, characterized in that: The blades of the wind turbine generator set are provided with a soundprint sensor, wherein the blade damage monitoring method comprises: During the rotation of the blades of the wind turbine generator set, a soundprint signal collected by a soundprint sensor on the blades is obtained; Extracting a voiceprint feature from the voiceprint signal, wherein a feature value of the voiceprint feature represents an energy level of the voiceprint signal; According to the rotation period of the blade, a characteristic value change of the voiceprint feature is detected to obtain a detection result, wherein the detection result represents the correlation between the characteristic value change of the voiceprint feature and the rotation period of the blade; According to the detection results, the blade damage monitoring result is obtained.

2. The blade damage monitoring method according to claim 1, characterized in that: The change of the characteristic value of the voiceprint feature is detected by the following method: Determine a feature value sequence of the voiceprint feature, wherein each feature value in the feature value sequence corresponds to a time series interval; Determine a plurality of peak intervals formed by the eigenvalues ​​in the eigenvalue sequence, wherein each peak interval contains an eigenvalue greater than a peak threshold; According to the rotation period of the blade, the characteristic value change between each peak interval is detected to obtain the detection result.

3. The blade damage monitoring method according to claim 2, characterized in that: The characteristic value changes between each peak interval are detected in the following way: Determine the representative value of the characteristic value in each peak interval; Based on the time values ​​corresponding to the representative values ​​of each peak interval, the target time series is obtained; Determine the difference between every two adjacent time values ​​in the target time series to obtain a plurality of time difference values; The detection result is obtained by comparing the duration of the rotation cycle of the blade with each time difference.

4. The blade damage monitoring method according to claim 3, characterized in that: The test results are obtained by: In response to the difference between each of the multiple time difference values ​​and the duration of the rotation period of the blade being less than a first preset value, a detection result is obtained indicating that there is a correlation between a change in a characteristic value of the voiceprint feature and the rotation period of the blade; In response to the difference between at least one of the multiple time difference values ​​and the duration of the rotation cycle of the blade being greater than or equal to the first preset value, the at least one time difference value is compared with the duration of the multiple rotation cycles to obtain the detection result.

5. The blade damage monitoring method according to claim 4, characterized in that: The comparing the at least one time difference with the durations of the plurality of rotation cycles to obtain the detection result comprises: In response to the fact that the difference between the value of each time difference value of the at least one time difference value divided into any number of parts and the duration of the rotation period is less than a second preset value, a detection result is obtained indicating that there is a correlation between the change in the characteristic value of the voiceprint feature and the rotation period of the blade; In response to the existence of a target time difference in the at least one time difference, a detection result is obtained that there is no correlation between the change in the characteristic value characterizing the voiceprint feature and the rotation period of the blade, wherein the difference between the value after the target time difference is divided into any number of parts and the duration of the rotation period is greater than or equal to a second preset value.

6. The blade damage monitoring method according to claim 2, characterized in that: The peak interval is determined by: Based on the eigenvalue sequence, determining a plurality of first eigenvalues ​​and a plurality of second eigenvalues, wherein the first eigenvalue is greater than the peak threshold and a previous eigenvalue of the first eigenvalue is less than or equal to the peak threshold, and the second eigenvalue is less than or equal to the peak threshold and a previous eigenvalue of the second eigenvalue is greater than the peak threshold; The peak interval is determined by determining a starting point of the peak interval from the multiple first eigenvalues ​​and determining an end point of the peak interval from the multiple second eigenvalues.

7. The blade damage monitoring method according to claim 6, characterized in that: The plurality of first eigenvalues ​​form a first sequence in time sequence, and the plurality of second eigenvalues ​​form a second sequence in time sequence, wherein the peak interval is further determined by: For the target feature value in the first sequence, the following operations are performed in sequence: Determine a target time interval based on the moment corresponding to the current target characteristic value and the duration of the rotation cycle of the blade, wherein the length of the target time interval is equal to the duration of the rotation cycle of the blade, and the target time interval includes the moment corresponding to the current target characteristic value; In the first sequence, searching for a first characteristic value that is within the target time interval and closest to the starting point of the target time interval to obtain a first target value; In the second sequence, searching for a second characteristic value that is within the target time interval and closest to an interval end point of the target time interval to obtain a second target value; The first target value is used as the interval starting point of the current peak interval, the second target value is used as the interval ending point of the current peak interval, and the first characteristic value after the first target value in the first sequence is used as the next target characteristic value, The initial target eigenvalue is the first eigenvalue in the first sequence.

8. The blade damage monitoring method according to claim 1, characterized in that: The step of obtaining a blade damage monitoring result according to the detection result includes: In response to the detection result indicating that there is a correlation between the change in the characteristic value of the voiceprint feature and the rotation period of the blade, and the number of characteristic values ​​of the voiceprint feature that are greater than a preset threshold is greater than or equal to a preset number, it is determined that the blade is at risk of damage.

9. The blade damage monitoring method according to claim 1, characterized in that: The voiceprint feature includes one of the following: In the frequency spectrum of the voiceprint signal, a statistical value of the amplitude of each frequency band interval in a plurality of frequency band intervals; Peak-to-peak value of the voiceprint signal; Sound pressure related quantity of voiceprint signal; The kurtosis of the voiceprint signal; The effective value of the voiceprint signal.

10. A blade damage monitoring system for a wind turbine generator set, characterized in that: The blade damage monitoring system comprises a monitoring platform and a soundprint sensor configured on the blades of the wind turbine generator set, wherein the monitoring platform comprises: processor; a memory for storing instructions executable by the processor, Wherein, when the processor executable instructions are executed by the processor, the processor is prompted to execute the blade damage monitoring method of a wind turbine generator set according to any one of claims 1 to 9.

11. The blade damage monitoring system according to claim 10, characterized in that: The wind turbine generator set also includes a main control system, and the monitoring platform sends the execution result of the processor to the main control system, so that the main control system performs a control action corresponding to the execution result, and the control action includes controlling the wind turbine generator set to retract blades and / or shut down.

12. A computer-readable storage medium, characterized in that: When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the blade damage monitoring method for a wind turbine generator set according to any one of claims 1 to 9.

13. A computer program product comprising computer executable instructions, characterized in that: When the computer executable instructions are executed by at least one processor, the blade damage monitoring method of a wind turbine generator set according to any one of claims 1 to 9 is implemented.

Images