Wind turbine generator blade damage monitoring method, device and equipment and storage medium

Abstract

The invention discloses a wind turbine generator blade damage monitoring method, device and equipment and a storage medium, which are applied to the field of blade damage monitoring, and the method comprises the following steps: acquiring blade signal data acquired by a calibrated multi-source sensor; identifying a leaf damage state represented in the characteristic signal of the leaf signal data based on time-frequency analysis; training a damage identification model based on a sample data set constructed based on a blade damage state; inputting to-be-analyzed blade signal data into the trained damage identification model to obtain a blade damage identification result; and executing a wind turbine generator control strategy based on a blade damage identification result and / or a modal analysis result. According to the method, the damage result is determined by performing time-frequency analysis on the leaf signal data acquired by the calibrated multi-source sensor, the sample set training model is constructed based on the damage result to perform real-time leaf damage identification, rapid identification of the damage result is realized through modal analysis, and the accuracy and the identification efficiency of damage identification are improved.

Description

Technical Field

[0001] The present invention relates to the field of blade damage monitoring, and in particular to a wind turbine blade damage monitoring method, a wind turbine blade damage monitoring device, an electronic device, and a computer-readable storage medium. Background Art

[0002] Currently, wind turbine blade health monitoring technology primarily focuses on monitoring the damage status of wind turbine blades. Current methods for blade damage monitoring primarily include vibration monitoring, acoustic emission monitoring, infrared thermal imaging, and video monitoring. While existing monitoring technologies can detect blade faults to a certain extent, they suffer from limitations such as limited monitoring parameters, which make it difficult to fully reflect blade health. The large volume of monitoring data makes analysis and processing difficult, hindering real-time fault diagnosis. Furthermore, sensors are susceptible to damage under complex operating conditions, leading to data distortion and potential misjudgment. Summary of the Invention

[0003] The purpose of the present invention is to provide a wind turbine blade damage monitoring method, device, equipment and storage medium, which are applied to the field of blade damage monitoring. The method determines the damage results by performing time-frequency analysis on blade signal data collected by calibrated multi-source sensors, constructs a sample set training model based on the damage results to perform real-time blade damage identification, and realizes rapid identification of damage results through modal analysis, thereby improving the accuracy and efficiency of damage identification.

[0004] To solve the above technical problems, the present invention provides a method for monitoring blade damage in a wind turbine generator system, comprising: Acquire blade signal data collected by the multi-source sensor after calibration; the multi-source sensor is arranged at different positions of the wind turbine blade; Extracting features from the blade signal data to obtain a characteristic signal, and identifying the blade damage state represented by the characteristic signal based on time-frequency analysis; Using the blade damage state as a label for the blade signal data to construct a sample data set, and training a damage recognition model based on the sample data set; The blade signal data to be analyzed is collected based on the multi-source sensor, and the blade signal data to be analyzed is input into the trained damage recognition model to obtain a blade damage recognition result; Modal analysis is performed based on the blade signal data to be analyzed to obtain a modal analysis result, and a wind turbine control strategy is executed based on the blade damage identification result and / or the modal analysis result.

[0005] Optionally, extracting features from the blade signal data to obtain feature data, and identifying the blade damage state represented by the feature data based on time-frequency analysis, includes: extracting time domain signal features from the blade signal data, and performing blade impact determination based on the time domain signal features; When it is determined that the blade has impacted based on the blade impact determination, performing frequency domain conversion on the blade signal data to obtain frequency domain signal data; A frequency domain signal feature is extracted from the frequency domain signal, and the blade damage state is determined based on the frequency domain signal feature.

[0006] Optionally, extracting frequency domain signal features from the frequency domain signal data includes: Performing high-frequency filtering on the frequency domain signal data to obtain the frequency domain signal data in a high frequency band; The frequency domain signal features are extracted from the frequency domain signal data in the high frequency band.

[0007] Optionally, performing modal analysis based on the blade signal data to be analyzed to obtain a modal analysis result includes: Performing modal decomposition based on blade acceleration data in the blade signal data to obtain a signal frequency of an n-th order mode; Determining a change in the natural frequency based on the signal frequency of the nth-order mode and the natural frequency of the blade in the initial state; Determining a stiffness change based on the natural frequency change and the nth-order modal vibration shape function; When the stiffness variation exceeds a preset threshold, the modal analysis result is determined to be blade fracture.

[0008] Optionally, obtaining blade signal data collected by the multi-source sensor after calibration includes: Obtaining the initial state stress value and initial state natural frequency of the wind turbine blade; determining a stress value based on ultrasonic data collected by an ultrasonic sensor in the multi-source sensor, and calibrating the ultrasonic sensor based on the stress value and the initial state stress value; Determining a signal frequency of an n-th order mode based on acceleration data collected by a MEMS acceleration sensor in the multi-source sensor, and calibrating the MEMS acceleration sensor based on the signal frequency of the n-th order mode and the natural frequency of the initial state; The blade signal data collected by the ultrasonic sensor and the MEMS acceleration sensor after calibration is obtained.

[0009] Optionally, the method further includes: Embedding a temperature and humidity sensor in the multi-source sensor to obtain temperature and humidity data collected by the temperature and humidity sensor; The blade signal data is compensated based on the temperature and humidity data.

[0010] Optionally, executing a wind turbine control strategy based on the blade damage identification result and / or modal analysis result includes: When the blade damage identification result is slight damage, the blade signal data is continuously recorded, and a monitoring report is generated every preset period; When the blade damage identification result is moderate damage, controlling the pitch angle to reduce the blade load; When the blade damage identification result is severe damage, controlling the wind turbine to stop running; When the modal analysis result indicates that the blade is broken, the wind turbine generator set is controlled to stop running.

[0011] In order to solve the above technical problems, the present invention provides a wind turbine blade damage monitoring device, comprising: The first module is used to obtain blade signal data collected by the multi-source sensor after calibration; the multi-source sensor is arranged at different positions of the wind turbine blade; The second module is used to extract features from the blade signal data to obtain a characteristic signal, and identify the blade damage state represented by the characteristic signal based on time-frequency analysis; A third module is configured to use the blade damage status as a label for the blade signal data to construct a sample data set, and train a damage recognition model based on the sample data set; A fourth module is configured to collect blade signal data to be analyzed based on the multi-source sensor, and input the blade signal data to be analyzed into the trained damage recognition model to obtain a blade damage recognition result; The fifth module is configured to perform modal analysis based on the blade signal data to be analyzed to obtain a modal analysis result, and execute a wind turbine control strategy based on the blade damage identification result and / or the modal analysis result.

[0012] To solve the above technical problems, the present invention provides an electronic device, comprising: Memory for storing computer programs; A processor is configured to implement the above-mentioned wind turbine blade damage monitoring method when executing the computer program.

[0013] In order to solve the above technical problems, the present invention provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, the above-mentioned wind turbine blade damage monitoring method is implemented.

[0014] It can be seen that the present invention obtains blade signal data collected by a multi-source sensor after calibration; the multi-source sensor is arranged at different positions of the wind turbine blade; the blade signal data is feature extracted to obtain a characteristic signal, and the blade damage state represented by the characteristic signal is identified based on time-frequency analysis; the blade damage state is used as a label of the blade signal data to construct a sample data set, and a damage recognition model is trained based on the sample data set; the blade signal data to be analyzed is collected based on the multi-source sensor, and the blade signal data to be analyzed is input into the trained damage recognition model to obtain a blade damage recognition result; modal analysis is performed on the blade signal data to be analyzed to obtain a modal analysis result, and the wind turbine control strategy is executed based on the blade damage recognition result and / or modal analysis result.

[0015] The present invention determines the damage results by performing time-frequency analysis on blade signal data collected by calibrated multi-source sensors, constructs a sample set training model based on the damage results to perform real-time blade damage identification, and realizes rapid identification of damage results through modal analysis, thereby improving the accuracy and efficiency of damage identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0017] Figure 1 A flow chart of a method for monitoring wind turbine blade damage provided by an embodiment of the present invention; Figure 2 This is an example diagram of normal blade data provided by an embodiment of the present invention; Figure 3 An example diagram of blade abnormality data provided by an embodiment of the present invention; Figure 4 This is a structural block diagram of a wind turbine blade damage monitoring device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] Currently, wind turbine blade health monitoring technologies primarily include vibration monitoring, acoustic emission monitoring, infrared thermal imaging, and video monitoring. While existing monitoring technologies can detect blade faults to a certain extent, they suffer from the following shortcomings: The limited monitoring parameters make it difficult to fully reflect blade health; monitoring equipment maintenance and replacement is complex; the volume of monitoring data is large, making analysis and processing difficult, making real-time fault diagnosis difficult; and sensors are susceptible to damage under complex operating conditions, resulting in data distortion and prone to misjudgment.

[0020] This paper proposes a wind turbine blade health monitoring method based on multi-sensor fusion + modal analysis + artificial intelligence diagnosis, which can accurately identify blade damage, reduce the misjudgment rate, and provide intelligent control strategies to improve the safety and reliability of wind turbines.

[0021] The following combination Figure 1 , Figure 1 A flow chart of a method for monitoring wind turbine blade damage provided by an embodiment of the present invention may include: S101: Acquire blade signal data collected by a calibrated multi-source sensor; the multi-source sensor is arranged at different positions of the wind turbine blade.

[0022] This embodiment can be deployed at multiple locations on each blade of a wind turbine to collect multi-source data from the wind turbine blades. This embodiment does not limit the specific type of multi-source sensor; it can generally include MEMS (Micro-Electro-Mechanical Systems Accelerometer) accelerometers, ultrasonic sensors, acoustic vibration sensors, and fiber optic sensors, with specific placement based on actual applications. This embodiment also does not limit the installation location of the multi-source sensor; it can generally be placed at the blade root, blade mid-section, and blade tip.

[0023] Since environmental changes may cause errors in the data collected by the sensor, this embodiment can embed a temperature and humidity sensor in the multi-source sensor to obtain the temperature and humidity data collected by the temperature and humidity sensor; and perform data compensation on the leaf signal data based on the temperature and humidity data.

[0024] After the multi-source sensors are arranged, this embodiment may calibrate the multi-source sensors to determine whether the multi-source sensors are in a normal operating state and whether the data collected by the sensors are reliable data.

[0025] This embodiment does not limit the specific method of calibrating the multi-source sensor. Taking the MEMS acceleration sensor and the ultrasonic sensor as an example, the initial state stress value and the initial state natural frequency of the wind turbine blade can generally be obtained; the stress value is determined based on the ultrasonic data collected by the ultrasonic sensor in the multi-source sensor, and the ultrasonic sensor is calibrated based on the stress value and the initial state stress value; the signal frequency of the nth order mode is determined based on the acceleration data collected by the MEMS acceleration sensor in the multi-source sensor, and the MEMS acceleration sensor is calibrated based on the signal frequency of the nth order mode and the initial state natural frequency; and the blade signal data collected by the ultrasonic sensor and the MEMS acceleration sensor after the calibration is completed is obtained.

[0026] In this embodiment, the MEMS accelerometer can be used as the core sensor in the multi-source sensor. The acceleration data collected by the MEMS accelerometer can effectively characterize the operating status of the blade. In the specific process of calibrating the MEMS accelerometer, the natural frequency of each blade in the initial state (that is, in a healthy and undamaged state) can be determined first, as shown in the following formula: ; Where, f n (x) is the natural frequency of the nth mode, K(x) is the stiffness of the blade at position x, and m is the unit mass.

[0027] The natural frequency of a wind turbine blade refers to the specific frequency of its free vibration. This is an inherent property of the blade, independent of external excitation. Natural frequency is a critical parameter in blade structural design, directly affecting its dynamic characteristics and reliability. A change in the blade's natural frequency indicates structural changes, indicating damage to the blade.

[0028] Furthermore, acceleration data collected by the MEMS acceleration sensor may be obtained, and modal analysis may be performed on the acceleration data to obtain the signal frequency of the acceleration data in the nth-order mode.

[0029] A frequency deviation between a signal frequency of an n-th order mode and a natural frequency of the n-th order mode is determined, and whether the MEMS acceleration sensor is abnormal is determined based on the frequency deviation.

[0030] Ideally, if the frequency deviation is not zero, the MEMS accelerometer is considered abnormal. However, in actual applications, errors may be introduced. Therefore, in this embodiment, when the frequency deviation is greater than a preset deviation threshold, the sensor is considered abnormal. When the frequency deviation is less than or equal to the preset deviation threshold, the sensor is considered normal, completing the sensor calibration.

[0031] S102: Extract features from the blade signal data to obtain a characteristic signal, and identify the blade damage state represented by the characteristic signal based on time-frequency analysis.

[0032] In this embodiment, the blade signal data may first be pre-processed, such as by wavelet transform, Kalman filter algorithm processing, etc., to achieve data noise reduction and improve data accuracy.

[0033] This embodiment can extract features from blade signal data to obtain a characteristic signal, and identify the blade damage status represented by the characteristic signal based on time-frequency analysis. Specifically, time domain signal features are extracted from the blade signal data, and blade impact determination is performed based on the time domain signal features. After the blade impact determination determines that a blade has experienced impact, the blade signal data is converted to the frequency domain to obtain frequency domain signal data. Frequency domain signal features are extracted from the frequency domain signal, and the blade damage status is determined based on the frequency domain signal features.

[0034] Furthermore, due to the blade damage, the signal in the high frequency band will have significant changes, such as Figure 2 and Figure 3 As shown, this embodiment can perform high-frequency filtering on the frequency domain signal data to extract the frequency domain signal data in the high frequency band, and extract the frequency domain signal features from the frequency domain signal data in the high frequency band.

[0035] In another embodiment, the frequency domain signal data can also be filtered to obtain frequency domain signal data in the low frequency band, medium frequency band and high frequency band respectively, and frequency domain analysis can be performed on the frequency domain signal data in each frequency band respectively, and the damage status of the blade can be jointly determined based on the frequency domain analysis results.

[0036] This embodiment does not limit the frequency range of each frequency band, which can generally be set based on actual applications. In this embodiment, the low frequency band can be a signal below 80HZ, the low frequency band can be a signal below 80HZ, the low frequency band can be a signal below 80HZ, the low frequency band can be a signal below 80HZ, the mid-frequency band can be a signal in the 320HZ~640HZ frequency band, and the high frequency band can be a signal in the 640HZ~1280HZ frequency band.

[0037] This embodiment does not limit the specific method of performing time-frequency analysis. Generally, in order to save computing resources, it is possible to first determine whether the blade is impacted through time domain analysis. After the blade is impacted, the time domain signal is converted into a frequency signal through Fourier transform, and the frequency domain signal characteristics are analyzed based on the frequency domain signal to determine the damage status of the blade.

[0038] Specifically, in the time domain, when the blade receives an impact, the signal has typical transient response characteristics: ; Where x(t) is the signal, A is the amplitude, e is the natural constant, β is the damping coefficient, ω is the natural frequency, and φ is the phase.

[0039] When a blade impacts, the amplitude A suddenly increases, and the signal energy E also increases instantaneously. Therefore, this embodiment can predetermine the baseline amplitude and baseline signal energy for the blade's initial state. The amplitude of the blade signal data is compared with the baseline amplitude to determine the amplitude mutation, while the energy mutation is determined by comparing the blade signal data's signal energy with the baseline signal energy.

[0040] In this embodiment, the impact of a blade can be determined by the amplitude mutation. Specifically, when the amplitude mutation exceeds the amplitude mutation threshold, the blade is determined to have been impacted. Alternatively, the impact of a blade can be determined by the energy mutation. Specifically, when the energy mutation exceeds the energy mutation threshold, the blade is determined to have been impacted. In practical applications, the approximate location of damage can be determined by analyzing signal energy. Therefore, in this embodiment, the energy mutation can be used for impact determination.

[0041] Furthermore, after completing the impact determination, this embodiment may perform frequency domain conversion on the blade signal data to obtain frequency domain signal data, for example, by performing conversion through fast Fourier transform: ; Where F(f) is the frequency domain signal after fast Fourier transform, f is the signal frequency, and t is time. When the blade breaks, the signal frequency increases significantly.

[0042] It can also be converted via short-time Fourier transform: ; Where STET(x(t)) is the frequency domain signal after short-time Fourier transform.

[0043] In this embodiment, frequency domain analysis can be performed by extracting spectrum information from the frequency domain signal after fast Fourier transformation, or by calculating spectrum energy from the frequency domain signal after short-time Fourier transformation.

[0044] Specifically, this embodiment can obtain experimental blade signal data of blades in different damage states in advance through experiments or simulations, and determine the reference spectrum energy in each damage state through the experimental blade signal data; wherein the reference spectrum energy can be an interval range.

[0045] In this embodiment, the calculated spectrum energy may be matched with the reference spectrum energy under each damage state, thereby determining the damage state corresponding to the collected blade signal data.

[0046] For example, this embodiment can determine the baseline spectrum energy intervals for mild damage state, moderate damage state and severe damage state, and determine the interval where the spectrum energy is located by matching the spectrum energy with the baseline spectrum energy interval under each damage state to determine the damage state of the blade signal data.

[0047] Furthermore, this embodiment can also obtain baseline spectrum information in a healthy state, extract the spectrum information from the frequency domain signal data, and determine the spectrum change based on the spectrum information and the baseline spectrum information. The spectrum change can then be used to determine whether the blade is broken. For example, when the spectrum change rate exceeds a spectrum change threshold, a blade breakage can be determined, and the corresponding wind turbine control strategy can be implemented.

[0048] In this embodiment, the matched damage state can be verified. The ultrasonic sensor can detect blade cracks and localized damage, so the matched damage state can be verified using the damage state represented by the ultrasonic data. If the verification fails, the corresponding blade signal data can be determined to be invalid.

[0049] Because blade signal data is derived from multiple sensor sources, this embodiment can perform damage state matching on each type of signal data and perform a joint analysis based on the damage state matching results for each type of signal data to determine the damage state of the blade signal data. Alternatively, a specific type of blade signal data, such as acceleration data collected by a MEMS accelerometer, can be selected from the multi-source sensor data, and damage state matching can be performed on the specific type of blade signal data, with the resulting damage state being used as the damage state of the blade signal data.

[0050] S103: Using the blade damage status as a label for the blade signal data to construct a sample data set, and training a damage recognition model based on the sample data set.

[0051] S104: collecting blade signal data to be analyzed based on multi-source sensors, and inputting the blade signal data to be analyzed into the trained damage recognition model to obtain a blade damage recognition result.

[0052] In this embodiment, a sample data set is constructed by using the blade damage state as a label of the blade signal data, and a damage recognition model is constructed. The damage recognition model is trained using the sample data set.

[0053] This embodiment does not limit the specific architecture of the damage identification model, which can be set based on actual applications. Generally, it can be a CNN (Convolutional Neural Networks) + LSTM (Long Short-Term Memory) architecture, and classification prediction is performed through random forest.

[0054] In this embodiment, the blade signal data to be analyzed acquired by the multi-source sensors can be input into the trained damage recognition model to obtain the blade damage recognition result.

[0055] In this embodiment, in order to improve computing efficiency, the collected data can be uploaded to the server through a data transmission device (such as a low-power LoRa (Long Range Radio) module or a 5G network). The server can perform data processing, time-frequency analysis, modal analysis, model training, damage identification, and control command issuance. The server can support edge computing.

[0056] S105: Performing modal analysis based on the blade signal data to be analyzed to obtain a modal analysis result, and executing a wind turbine control strategy based on the blade damage identification result and / or the modal analysis result.

[0057] In this embodiment, modal analysis can be performed based on the blade signal data to be analyzed to obtain modal analysis results, and the control strategy of the wind turbine can be jointly executed through the modal analysis results and the damage identification results.

[0058] This embodiment does not limit the specific method of performing modal analysis. Generally, modal decomposition can be performed based on the blade acceleration data in the blade signal data to obtain the signal frequency of the nth-order mode; the natural frequency change is determined based on the signal frequency of the nth-order mode and the natural frequency of the blade in the initial state; the stiffness change is determined based on the natural frequency change and the nth-order modal vibration shape function; when the stiffness change exceeds a preset threshold, the modal analysis result is determined to be a blade fracture.

[0059] Specifically, local damage causes the blade stiffness to decrease, and the stiffness change at the damage point can be calculated using the modal strain energy theory: ; Where, K (x) is the stiffness variation, n is the modal order, f n is the change of natural frequency at the damage point, n (x) is the nth-order mode shape function.

[0060] When the stiffness variation calculated by modal analysis is greater than the stiffness variation threshold, it can be determined that the blade is broken, that is, the modal analysis result is blade fracture.

[0061] When a blade breaks, the wind turbine needs to be controlled in a timely manner to reduce the probability of an accident. The timeliness of model identification is low. Therefore, this embodiment can introduce a wind turbine control strategy based on modal analysis results on the basis of executing the wind turbine control strategy based on the blade damage identification results.

[0062] Specifically, when the blade damage identification result is mild damage, the blade signal data is continuously recorded and a monitoring report is generated every preset period; when the blade damage identification result is moderate damage, the pitch angle is controlled to reduce the blade load; when the blade damage identification result is severe damage, the wind turbine is controlled to stop running; when the modal analysis result is a blade breakage, the wind turbine is controlled to stop running.

[0063] In this embodiment, the damage location can be determined. The specific principle is that when a blade is damaged, the fluctuation of blade signal data collected by sensors closer to the blade is greater than the fluctuation of blade signal data collected by sensors closer to the blade. In this embodiment, the fracture location can be determined based on the change in the signal frequency of the nth-order mode in the acceleration data collected by the sensor compared to the natural frequency. A larger change indicates that the fracture location is closer to the sensor collecting the corresponding acceleration data.

[0064] Specifically, this embodiment can construct characteristic baseline values for each sensor location on a healthy blade. For example, for sensors installed at the root, tip, and mid-blade, characteristic baseline values F1, F2, and F3, as well as corresponding damage thresholds K1, K2, and K3, can be determined for each sensor installation location. The damage threshold is the maximum characteristic value that can be captured at a sensor location. The closer the characteristic value captured by a sensor is to the damage threshold, the closer the damage is to that sensor location. Due to the structural characteristics of the blade, the damage threshold may not be the same at each location.

[0065] In this embodiment, the characteristic deviation can be determined by the characteristic value of the signal at each position acquired in real time. The larger the characteristic deviation is, the closer the damage is to the position.

[0066] For example, the baseline characteristic values and damage thresholds at each position are determined: Position 1 (blade root): F1=100, K1=150; Position 2 (leaf center): F2=80, K2=130; Position 3 (blade tip): F3=60, K3=100.

[0067] In a certain monitoring data, the characteristic values of the collected characteristic data are: (leaf root) T1=110, (leaf middle) T2=85, (leaf tip) T3=90.

[0068] At this point the absolute deviation of the feature can be calculated: ΔF1=110-100=10; ΔF2=85-80=5; ΔF3=90-60=30; By comparing, we find that ΔF3 is the maximum value, which means that the maximum deviation is at position 3, and therefore the damage is near position 3 (the blade tip). At the same time, the characteristic value (90) of position 3 is close to the damage threshold (100), while positions 1 and 2 are far below their respective thresholds, which also confirms that damage may have occurred near position 3.

[0069] Based on the above embodiments, the present invention determines the damage results by performing time-frequency analysis on the blade signal data collected by the calibrated multi-source sensor, constructs a sample set training model based on the damage results to perform real-time blade damage identification, and realizes rapid identification of damage results through modal analysis, thereby improving the accuracy and efficiency of damage identification.

[0070] The following combination Figure 3 , FIG is a flow chart of a wind turbine blade damage monitoring device provided by an embodiment of the present invention, the device may include: The first module 100 is used to obtain blade signal data collected by a multi-source sensor after calibration; the multi-source sensor is arranged at different positions of the wind turbine blade; The second module 200 is configured to extract features from the blade signal data to obtain a characteristic signal, and identify the blade damage state represented by the characteristic signal based on time-frequency analysis; The third module 300 is configured to use the blade damage state as a label of the blade signal data to construct a sample data set, and train a damage recognition model based on the sample data set; The fourth module 400 is configured to collect blade signal data to be analyzed based on the multi-source sensor, and input the blade signal data to be analyzed into the trained damage recognition model to obtain a blade damage recognition result; The fifth module 500 is configured to perform a modal analysis based on the blade signal data to be analyzed to obtain a modal analysis result, and execute a wind turbine control strategy based on the blade damage identification result and / or the modal analysis result.

[0071] Based on the above embodiments, the present invention determines the damage results by performing time-frequency analysis on the blade signal data collected by the calibrated multi-source sensor, constructs a sample set training model based on the damage results to perform real-time blade damage identification, and realizes rapid identification of damage results through modal analysis, thereby improving the accuracy and efficiency of damage identification.

[0072] Based on the above embodiment, the second module 200 may include: A first unit is configured to extract time domain signal features from the blade signal data and perform blade impact determination based on the time domain signal features; A second unit is configured to, when it is determined based on the blade impact determination that the blade has impacted, perform frequency domain conversion on the blade signal data to obtain frequency domain signal data; The third unit is configured to extract frequency domain signal features from the frequency domain signal, and determine the damage status of the blade based on the frequency domain signal features.

[0073] Based on the above embodiments, the second unit may include: A first subunit is configured to perform high-frequency filtering on the frequency domain signal data to obtain the frequency domain signal data in a high frequency band; The second subunit is configured to extract the frequency domain signal features from the frequency domain signal data in the high frequency band.

[0074] Based on the above embodiments, the fifth module 500 may include: a fourth unit, configured to perform modal decomposition based on blade acceleration data in the blade signal data to obtain a signal frequency of an n-th order mode; A fifth unit is configured to determine a change in the natural frequency based on the signal frequency of the n-th order mode and the natural frequency of the blade in an initial state; A sixth unit is used to determine a stiffness change based on the natural frequency change and the nth-order mode shape function; The seventh unit is configured to determine that the modal analysis result is blade fracture when the stiffness change exceeds a preset threshold.

[0075] Based on the above embodiments, the first module 100 may include: An eighth unit is used to obtain the initial state stress value and initial state natural frequency of the wind turbine blade; a ninth unit, configured to determine a stress value based on ultrasonic data collected by an ultrasonic sensor in the multi-source sensor, and calibrate the ultrasonic sensor based on the stress value and the initial state stress value; a tenth unit, configured to determine a signal frequency of an n-th order mode based on acceleration data collected by a MEMS acceleration sensor in the multi-source sensor, and calibrate the MEMS acceleration sensor based on the signal frequency of the n-th order mode and the natural frequency of the initial state; The eleventh unit is used to obtain the blade signal data collected by the ultrasonic sensor and the MEMS acceleration sensor after calibration.

[0076] Based on the above embodiments, the present invention further includes: A sixth module is configured to embed a temperature and humidity sensor in the multi-source sensor and obtain temperature and humidity data collected by the temperature and humidity sensor; The seventh module is used to perform data compensation on the blade signal data based on the temperature and humidity data.

[0077] Based on the above embodiments, executing a wind turbine control strategy based on the blade damage identification result includes: The eleventh unit is configured to continuously record the blade signal data and generate a monitoring report at a preset period when the blade damage identification result is slight damage; The twelfth unit is configured to control the pitch angle to reduce the blade load when the blade damage identification result is moderate damage; The thirteenth unit is configured to control the wind turbine to stop running when the blade damage identification result is severe damage; The fourteenth unit is configured to control the wind turbine to stop running when the modal analysis result indicates that the blade is broken.

[0078] Based on the above embodiments, the present invention further provides an electronic device, which may include a memory and a processor. The memory stores a computer program, and the processor, when invoking the computer program in the memory, can implement the steps provided in the above embodiments. Of course, the device may also include various necessary network interfaces, a power supply, and other components.

[0079] The present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by an execution terminal or a processor, the method provided in the embodiment of the present invention can be implemented. The storage medium may include: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc., various media that can store program codes.

[0080] In this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus comprising the element.

Claims

1. A wind turbine blade damage monitoring method, characterized in that: include: Acquire blade signal data collected by the multi-source sensor after calibration; The multi-source sensors are arranged at different positions of the wind turbine blades; Extracting features from the blade signal data to obtain a characteristic signal, and identifying the blade damage state represented by the characteristic signal based on time-frequency analysis; Using the blade damage state as a label for the blade signal data to construct a sample data set, and training a damage recognition model based on the sample data set; The blade signal data to be analyzed is collected based on the multi-source sensor, and the blade signal data to be analyzed is input into the trained damage recognition model to obtain a blade damage recognition result; Modal analysis is performed based on the blade signal data to be analyzed to obtain a modal analysis result, and a wind turbine control strategy is executed based on the blade damage identification result and / or the modal analysis result.

2. The wind turbine blade damage monitoring method according to claim 1, characterized in that: Extracting features from the blade signal data to obtain feature data, and identifying the blade damage state represented by the feature data based on time-frequency analysis, including: extracting time domain signal features from the blade signal data, and performing blade impact determination based on the time domain signal features; When it is determined that the blade has impacted based on the blade impact determination, performing frequency domain conversion on the blade signal data to obtain frequency domain signal data; A frequency domain signal feature is extracted from the frequency domain signal, and the blade damage state is determined based on the frequency domain signal feature.

3. The wind turbine blade damage monitoring method according to claim 2, characterized in that: Extracting frequency domain signal features from the frequency domain signal data includes: Performing high-frequency filtering on the frequency domain signal data to obtain the frequency domain signal data in a high frequency band; The frequency domain signal features are extracted from the frequency domain signal data in the high frequency band.

4. The wind turbine blade damage monitoring method according to claim 1, characterized in that: Performing modal analysis based on the blade signal data to be analyzed to obtain modal analysis results includes: Performing modal decomposition based on blade acceleration data in the blade signal data to obtain a signal frequency of an n-th order mode; Determining a change in the natural frequency based on the signal frequency of the nth-order mode and the natural frequency of the blade in the initial state; Determining a stiffness change based on the natural frequency change and the nth-order modal vibration shape function; When the stiffness variation exceeds a preset threshold, the modal analysis result is determined to be blade fracture.

5. The wind turbine blade damage monitoring method according to claim 1, characterized in that: Obtain blade signal data collected by the calibrated multi-source sensor, including: Obtaining the initial state stress value and initial state natural frequency of the wind turbine blade; determining a stress value based on ultrasonic data collected by an ultrasonic sensor in the multi-source sensor, and calibrating the ultrasonic sensor based on the stress value and the initial state stress value; Determining a signal frequency of an n-th order mode based on acceleration data collected by a MEMS acceleration sensor in the multi-source sensor, and calibrating the MEMS acceleration sensor based on the signal frequency of the n-th order mode and the natural frequency of the initial state; The blade signal data collected by the ultrasonic sensor and the MEMS acceleration sensor after calibration is obtained.

6. The wind turbine blade damage monitoring method according to claim 1, characterized in that: Also includes: Embedding a temperature and humidity sensor in the multi-source sensor to obtain temperature and humidity data collected by the temperature and humidity sensor; The blade signal data is compensated based on the temperature and humidity data.

7. The wind turbine blade damage monitoring method according to claim 1, characterized in that: Executing a wind turbine control strategy based on the blade damage identification result and / or the modal analysis result includes: When the blade damage identification result is slight damage, the blade signal data is continuously recorded, and a monitoring report is generated every preset period; When the blade damage identification result is moderate damage, controlling the pitch angle to reduce the blade load; When the blade damage identification result is severe damage, controlling the wind turbine to stop running; When the modal analysis result indicates that the blade is broken, the wind turbine generator set is controlled to stop running.

8. A wind turbine blade damage monitoring device, characterized in that: include: The first module is used to obtain blade signal data collected by the multi-source sensor after calibration; The multi-source sensors are arranged at different positions of the wind turbine blades; The second module is used to extract features from the blade signal data to obtain a characteristic signal, and identify the blade damage state represented by the characteristic signal based on time-frequency analysis; A third module is configured to use the blade damage status as a label for the blade signal data to construct a sample data set, and train a damage recognition model based on the sample data set; A fourth module is configured to collect blade signal data to be analyzed based on the multi-source sensor, and input the blade signal data to be analyzed into the trained damage recognition model to obtain a blade damage recognition result; The fifth module is configured to perform modal analysis based on the blade signal data to be analyzed to obtain a modal analysis result, and execute a wind turbine control strategy based on the blade damage identification result and / or the modal analysis result.

9. An electronic device, characterized in that: include: Memory for storing computer programs; A processor is configured to implement the wind turbine blade damage monitoring method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, the wind turbine blade damage monitoring method according to any one of claims 1 to 7 is implemented.

Images