A wind turbine fault early warning system based on big data

Abstract

The present application relates to the technical field of wind turbine fault early warning, and particularly relates to a wind turbine fault early warning system based on big data. Through the cooperative matching of the spanwise segmentation module, the region selection module, the risk analysis module, the spread analysis module and the early warning pushing module, the spanwise segmentation module determines the segmentation position point based on the spanwise mistuning degree as the space segmentation basis, the region selection module determines the blade section between adjacent segmentation position points as a suspicious sub-region, and the risk analysis module determines the risk tendency based on the trailing edge angle feature and the cyclone instability feature, so as to separate the asymmetric degradation from the sensor drift, the spread analysis module determines the spread distribution range based on the change trend of the trailing edge angle of the suspicious sub-region with the risk tendency, and tracks the spatial expansion process of the degradation, so as to improve the perception ability of the early warning system to the early structural degradation and the accuracy of the early warning.

Description

Technical Field

[0001] This invention relates to the field of wind turbine fault early warning technology, and in particular to a wind turbine fault early warning system based on big data. Background Technology

[0002] With the increasing size of wind turbines and the large-scale development of offshore wind power, the operating environment of blades is becoming increasingly harsh. The trailing edge of the blade is subjected to wind and rain erosion, ultraviolet radiation and salt spray corrosion for a long time. The aerodynamic integrity of the blade is closely related to the safe operation of the unit. The aerodynamic noise characteristics generated by trailing edge vortex shedding are closely related to the degradation state of the blade surface material. Blade condition monitoring faces a fundamental challenge in identifying early structural degradation. This capability has become an indispensable part of the wind power predictive maintenance system.

[0003] Chinese Patent Publication No. CN114135449A discloses a method for early warning of wind turbine blade faults, relating to the field of wind turbine generators. It aims to improve the difficulty in assessing internal damage and faults in wind turbine blades. The method includes obtaining a first score characterizing wind speed power, a second score characterizing wind speed hub speed, and a third score characterizing wind speed tip speed ratio based on multiple sets of wind speed power, multiple sets of wind speed hub speed, and multiple sets of wind speed tip speed ratio; obtaining a fourth score characterizing axial vibration data at the second-by-second level based on multiple sets of axial vibration data; and determining the state of the wind turbine blades based on the first, second, third, and fourth scores. By obtaining the first, second, third, and fourth scores based on wind speed power, wind speed hub speed, wind speed tip speed ratio, and axial vibration data at the second-by-second level, and then determining the state of the wind turbine blades based on these scores, the assessment difficulty is reduced.

[0004] However, the following problems still exist in the existing technology. In wind turbine blade fault early warning, the gradual decrease in vortex shedding intensity caused by asymmetric degradation of the blade trailing edge is misjudged as sensor drift and filtered out by the existing system, thereby reducing the early warning system's ability to detect early structural degradation. Summary of the Invention

[0005] To address this issue, the present invention provides a wind turbine fault early warning system based on big data, which overcomes the problem in the prior art where the gradual decrease in vortex shedding intensity caused by asymmetric degradation of the blade trailing edge is misjudged as sensor drift and filtered out, thereby reducing the early warning system's ability to detect early structural degradation.

[0006] To achieve the above objectives, the present invention provides a wind turbine fault early warning system based on big data, comprising, The spanwise segmentation module is used to determine the segmentation position of each blade based on the spanwise mistuning degree; wherein, the spanwise mistuning degree is determined based on the arc length of the upper surface and the arc length of the lower surface at the same spanwise position of the blade. The region selection module is used to determine suspicious sub-regions based on the positional relationship of the segmentation points; The risk analysis module is used to determine the risk tendency of suspicious sub-regions based on trailing edge angle characteristics and cyclone instability characteristics. The trailing edge angle characteristics are determined based on the angle between the trailing edge tangent of the upper arc surface and the trailing edge tangent of the lower surface at the same spanwise position, and the cyclone instability characteristics are determined based on the peak time difference of the vortex shedding signal at adjacent spanwise positions. The spread analysis module is used to determine the spread distribution range based on the changing trend of the trailing edge angle characteristics of suspicious sub-regions with risk potential; The early warning push module is used to push early warnings based on the spread and distribution range.

[0007] Furthermore, the extension segmentation module includes, The spanwise mistuning analysis sub-unit is used to determine the spanwise mistuning degree based on the ratio of the upper surface arc length to the lower surface arc length at the same spanwise position of the blade.

[0008] Furthermore, the extension segmentation module also includes, The segmentation location determination sub-element is used to determine the segmentation location point where the difference between the spanwise mistuning degree and the reference mistuning degree exceeds the reference deviation value.

[0009] Furthermore, the region selection module includes, Sub-regions are defined as sub-units, which are used to identify the blade region enclosed by the upper and lower surfaces between adjacent segmentation points as a suspicious sub-region.

[0010] Furthermore, the risk analysis module includes, The risk tendency determination subunit is used to determine that the suspected sub-region has a risk tendency when the trailing edge angle feature of the suspected sub-region is lower than the reference trailing edge angle value and the cyclone instability feature exceeds the reference cyclone instability value.

[0011] Furthermore, the risk analysis module also includes, The included angle feature calculation subunit is used to determine the mean value of the included angle between the trailing edge tangent of the upper arc surface and the trailing edge tangent of the lower surface at the same spanwise position within the monitoring period as the included angle feature.

[0012] Furthermore, the risk analysis module also includes, The instability characteristic calculation subunit is used to determine the mean of the peak time difference within the monitoring period as the cyclone instability characteristic; wherein, the peak time difference is determined based on the time interval between the peak times at adjacent measuring points, and the peak time is the moment when the amplitude of the vortex shedding signal exceeds the reference amplitude.

[0013] Furthermore, the spanwise interval corresponding to the measuring point contains several spanwise positions.

[0014] Furthermore, the propagation analysis module includes, The spread range delineation sub-unit is used to define the spread distribution range as the longitudinal area between the spread start position and the spread end position.

[0015] Furthermore, the propagation initiation position is the position where the trailing edge angle feature first shows a unidirectional decrease along the span, and the propagation termination position is the position where the trailing edge angle feature shows a reverse increase along the span.

[0016] Compared with existing technologies, this invention improves the early warning system's ability to perceive early structural degradation and enhances the accuracy of early warning by employing a spanwise segmentation module, a region selection module, a risk analysis module, a propagation analysis module, and an early warning push module. The spanwise segmentation module uses spanwise mistuning as the basis for spatial segmentation to determine the segmentation location points. The region selection module identifies blade segments between adjacent segmentation locations as suspicious sub-regions. The risk analysis module determines the risk tendency based on trailing edge angle characteristics and cyclone instability characteristics, thereby separating asymmetric degradation from sensor drift. The propagation analysis module determines the propagation distribution range based on the changing trend of the trailing edge angle of suspicious sub-regions with risk tendency, tracking the spatial expansion process of degradation.

[0017] In particular, this invention determines the segmentation points of each blade based on spanwise detuning, and identifies suspicious sub-regions based on the positional relationships of these segmentation points. In actual wind turbine operation scenarios, common-mode interference such as sensor drift and ambient temperature changes has a uniform impact on all spanwise positions of the entire blade. Monitoring signals at all spanwise positions rise and fall synchronously, without local abrupt changes at specific spanwise positions. By calculating the ratio of the arc length of the upper surface to the arc length of the lower surface at all spanwise positions of the blade as the spanwise detuning, the segmentation points are determined. The blade segment enclosed by the upper and lower surfaces between adjacent segmentation points is identified as a suspicious sub-region. The spanwise detuning is directly extracted from the measured data of the blade cross-sectional profile. Sensor drift only changes the amplitude of the electronic signal and does not alter the physical dimensions of the blade itself, thus eliminating the interference of sensor drift on spatial segmentation at the data source. The segmentation method based on ratio abrupt changes ensures that anomalies occurring in local areas at the blade tip are not masked by the global average of the entire blade, eliminating the submergence effect of common-mode interference on local structural degradation. Different blades have different service life, windward orientation, and history of wind and rain erosion, resulting in different spanwise distributions of material loss on their upper surface. Consequently, the segmentation points and the division methods for suspicious sub-regions also differ. This enables spatial segmentation to match the actual degradation state of each blade, separating local degradation from global drift from a spatial dimension. This eliminates the risk of misjudgment that traditional methods might incorrectly classify local degradation signals as global drift and systematically filter them out, thereby improving the reliability of the early warning system in identifying local structural degradation.

[0018] In particular, this invention determines the risk tendency of suspicious sub-regions based on trailing edge angle characteristics and cyclone instability characteristics. In actual wind turbine operation scenarios, the upper surface of the trailing edge of the blade is preferentially subjected to wind and rain erosion and ultraviolet radiation for a long time. After the coating peels off, the substrate erosion develops non-uniformly along the spanwise direction, directly changing the geometric angle between the trailing edge tangent of the upper arc surface and the trailing edge tangent of the lower surface at the corresponding spanwise position. The average value of this angle during the monitoring period, i.e., the trailing edge angle characteristic, deviates from the design reference accordingly. At the same time, the spanwise non-uniform distribution of material loss on the upper and lower surfaces directly disrupts the originally orderly alternating shedding process of the trailing edge vortex street. Once the arcuate relationship at the trailing edges of the upper and lower surfaces becomes unbalanced due to material loss, the timing and manner of airflow separation on both sides no longer align. The vortex shedding rhythm becomes disordered, and the time intervals between noise peaks at adjacent spanwise positions become erratic. The cyclonic instability characteristics, based on the statistical mean of the time differences between peaks, increase. Therefore, asymmetric degradation is verified from both geometric profile and aerodynamic characteristics. The nature of asymmetric degradation determines that the trailing edge angle characteristics and cyclonic instability characteristics deteriorate simultaneously. The trailing edge angle, as a geometric quantity, is directly extracted from the measured data of the blade cross-sectional profile. Sensor drift can only change the amplitude of the electronic signal but cannot reduce the geometric angle measured by laser scanning. The cyclonic instability characteristics, as a temporal quantity, are directly extracted from the peak moment of the vortex shedding signal. Using a trailing edge angle characteristic lower than the baseline trailing edge angle value and cyclonic instability characteristics exceeding the baseline cyclonic instability value as conditions for determining the potential risk tendency of suspicious sub-regions, asymmetric degradation is verified from both geometric profile and aerodynamic characteristics, ensuring accurate capture of anomalies caused by material loss. Common-mode interference such as sensor drift is naturally eliminated because it is impossible to simultaneously produce a phenomenon of decreasing geometric quantity and increasing temporal quantity. This removes asymmetric degradation from sensor drift and improves the reliability of the early warning system in perceiving early structural degradation.

[0019] In particular, this invention determines the spread distribution range based on the changing trend of the trailing edge angle characteristics of suspected sub-regions with risk tendencies. In actual wind turbine operation scenarios, once trailing edge material loss occurs in a certain local section, the continuous effects of wind and rain erosion and aerodynamic loads will drive the degradation boundary to expand longitudinally. By using the spatial gradient change law of the trailing edge angle characteristics to track the degradation front, the deviation of the cross-sectional geometric angle decreases as it transitions from the degradation zone to the non-degraded zone, and the trailing edge angle characteristics decrease unidirectionally along the longitudinal direction. The starting point of this trend is determined as the spread initiation position; when approaching the boundary between the degradation boundary and healthy material, the trailing edge angle characteristics increase in the opposite direction, and this position is determined as the spread termination position. The spread analysis module delineates the longitudinal section between the two as the spread distribution range, eliminating the problem of fragmented warning range caused by risk tendency determination only being able to identify discrete risk points and not be able to determine the actual range of degradation. The spread distribution range connects adjacent risk points in the extension direction into continuous segments of the same degradation event. Based on this, the early warning push module provides a one-time early warning for the entire segment, providing a clear repair boundary for operation and maintenance decisions. This improves the accuracy of the early warning range and the actual damaged segment of the blade, as well as the reliability of the early warning system's perception of early structural degradation. Attached Figure Description

[0020] Figure 1 This is a system block diagram of a wind turbine fault early warning system based on big data, as an embodiment of the invention. Figure 2 This is a schematic diagram illustrating the steps of the big data-based wind turbine fault early warning method according to an embodiment of the invention; Figure 3 A logic diagram for determining the segmentation position points of each blade in an embodiment of the invention; Figure 4 This is a logic diagram for determining the risk tendency of suspicious sub-regions in an embodiment of the invention. Detailed Implementation

[0021] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0022] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0023] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the term "connection" should be interpreted broadly. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0024] Please see Figure 1 The diagram shown is a system block diagram of a wind turbine fault early warning system based on big data, according to an embodiment of the invention. The system of the present invention includes: The spanwise segmentation module is used to determine the segmentation position of each blade based on the spanwise mistunt. The region selection module is used to determine suspicious sub-regions based on the positional relationship of the segmentation points; The risk analysis module is used to determine the risk tendency of suspicious sub-regions based on trailing edge angle characteristics and cyclone instability characteristics. The spread analysis module is used to determine the spread distribution range based on the changing trend of the trailing edge angle characteristics of suspicious sub-regions with risk potential; The early warning push module is used to push early warnings based on the spread and distribution range.

[0025] Please see Figure 2 The diagram illustrates the steps of a wind turbine fault early warning method based on big data, as described in this invention. The method includes: The splitting point of each blade is determined based on the spanwise mistuning. Suspicious sub-regions are determined based on the positional relationships of the segmentation points; The risk tendency of suspicious sub-regions is determined based on the trailing edge angle characteristics and cyclone instability characteristics. The spread distribution range is determined based on the changing trend of the trailing edge angle characteristics of suspicious sub-regions with risk potential; Early warnings will be sent based on the extent of the spread.

[0026] Please see Figure 3 As shown, this is a logic decision diagram for determining the segmentation position points of each blade according to an embodiment of the invention. The spanwise segmentation module described in this embodiment includes... The spanwise mistuning analysis sub-unit is used to determine the spanwise mistuning degree based on the ratio of the upper surface arc length to the lower surface arc length at the same spanwise position of the blade.

[0027] In this embodiment, the cross-sectional profile is cut out position by position along the blade spanwise. On the cross-sectional profile corresponding to each spanwise position, the arc length of the upper surface and the arc length of the lower surface are obtained and determined as the spanwise mistunt.

[0028] It is understandable that the arc lengths of the upper and lower surfaces can be determined by laser scanning or by non-contact geometric measurement methods such as photogrammetry from the blade cross-sectional profile. As long as the geometric dimensions of the blade cross-sectional profile can be accurately obtained, the specific measurement method can be flexibly selected according to the site conditions, and will not be elaborated further here.

[0029] Specifically, the spanwise segmentation module also includes, The segmentation location determination sub-element is used to determine the segmentation location point where the difference between the spanwise mistuning degree and the reference mistuning degree exceeds the reference deviation value.

[0030] In this embodiment, the reference mistuning degree is the standard ratio of the arc length of the upper surface to the arc length of the lower surface at each spanwise position of the blade at the time of manufacture. As an example, those skilled in the art can select the blade's manufacturing inspection or final acceptance stage, obtain the cross-sectional profile corresponding to each spanwise position through laser scanning or photogrammetry, extract the arc length of the upper surface and the arc length of the lower surface, calculate the ratio between the two, and use the ratio corresponding to each spanwise position as the reference mistuning degree. The specific method of obtaining this ratio does not constitute a limitation of the present invention and will not be elaborated here.

[0031] In this embodiment, the reference deviation value is the normal fluctuation range of the spanwise mistuning at each spanwise position of the blade in its factory condition. As an example, those skilled in the art can select the spanwise mistuning of multiple blades from the same batch at the same spanwise position and determine the maximum difference between each blade as the reference deviation value. The specific method of obtaining this value does not constitute a limitation of the present invention and will not be elaborated here.

[0032] In actual wind turbine operation scenarios, common-mode interference such as sensor drift and ambient temperature changes has a uniform impact on all spanwise positions of the blade. Monitoring signals at all spanwise positions rise and fall synchronously, without local abrupt changes at specific spanwise positions. In this embodiment, the segmentation position is determined by calculating the ratio of the upper surface arc length to the lower surface arc length at all spanwise positions of the blade as the spanwise detuning degree. The spanwise detuning degree is directly extracted from the measured data of the blade cross-sectional profile. Sensor drift can only change the amplitude of the electronic signal and cannot change the physical dimensions of the blade itself, thus eliminating the interference of sensor drift on spatial segmentation from the data source. The segmentation method based on ratio abrupt changes ensures that anomalies occurring in local areas at the blade tip are not masked by the global average of the entire blade, eliminating the submergence effect of common-mode interference on local structural degradation. This improves the accuracy of spatial segmentation in identifying local structural degradation and determining suspicious sub-regions.

[0033] Specifically, the region selection module includes, Sub-regions are defined as sub-units, which are used to identify the blade region enclosed by the upper and lower surfaces between adjacent segmentation points as a suspicious sub-region.

[0034] In this embodiment, within the spanwise interval between adjacent segmentation points, the upper and lower surfaces together enclose a complete blade segment. The spanwise detuning at each spanwise position within this segment deviates from the reference, indicating that the cross-section of this segment has deviated as a whole. This entire segment is designated as a suspicious sub-region, serving as the basic spatial unit for the risk analysis module to extract and determine trailing edge angle features and cyclone instability features; further details are omitted here.

[0035] Please see Figure 4 As shown, it is a logical determination diagram for identifying the risk tendency of suspicious sub-regions in an embodiment of the invention. The risk analysis module described in this embodiment includes, The risk tendency determination subunit is used to determine that the suspected sub-region has a risk tendency when the trailing edge angle feature of the suspected sub-region is lower than the reference trailing edge angle value and the cyclone instability feature exceeds the reference cyclone instability value.

[0036] In this embodiment, the reference trailing edge angle value is a reference value for the angle between the trailing edge tangent of the upper arc surface and the trailing edge tangent of the lower surface at each spanwise position when the blade leaves the factory. As an example, during the blade's factory quality inspection stage, those skilled in the art can select the cross-sectional profiles of multiple blades from the same batch at the same spanwise position, measure the trailing edge angle of each blade at that spanwise position, and use the average of the trailing edge angles of all blades as the reference trailing edge angle value at that spanwise position. The reference trailing edge angle value reflects the actual manufacturing average level of the trailing edge angle of the same batch of blades in the factory state. The specific method of obtaining it does not constitute a limitation of the present invention and will not be elaborated here.

[0037] In this embodiment, the reference cyclone instability value is the normal level of the peak time difference of vortex shedding signals between adjacent cross sections at each spanwise position under healthy operating conditions after the blade leaves the factory. As an example, those skilled in the art can collect vortex shedding signals at each spanwise position of the blade's trailing edge during the initial operation phase of the wind turbine, before the blade shows signs of degradation, extract the peak time of each vortex shedding cycle between adjacent spanwise positions, calculate the time difference, and use the average of the peak time differences within this healthy operating phase as the reference cyclone instability value for that adjacent spanwise position. The reference cyclone instability value reflects the level of vortex shedding synchronicity of the blade under factory conditions; its specific acquisition method does not constitute a limitation of the present invention and will not be elaborated further.

[0038] Specifically, the risk analysis module also includes, The included angle feature calculation subunit is used to determine the mean value of the included angle between the trailing edge tangent of the upper arc surface and the trailing edge tangent of the lower surface at the same spanwise position within the monitoring period as the included angle feature.

[0039] In this embodiment, the monitoring period is the cumulative duration of several rotations of the wind turbine blades under normal operating conditions. As an example, those skilled in the art can select a period when the blades are operating stably near the rated wind speed and continuously collect the trailing edge angle characteristics. The duration of the monitoring period is designed to ensure that it covers the cross-sectional geometry of the blades at different rotation phase angles and eliminates random errors in single measurements. The specific duration can be flexibly set according to the operating conditions of the wind turbines and the accuracy of the measuring equipment, which will not be elaborated further here.

[0040] In actual wind turbine operation scenarios, the upper surface of the trailing edge of the blade is preferentially subjected to wind and rain erosion and ultraviolet radiation over a long period. After the coating peels off, the substrate erosion develops non-uniformly along the spanwise direction, directly changing the geometric angle between the trailing edge tangents of the upper and lower surfaces at the corresponding spanwise position. The average value of this angle during the monitoring period, i.e., the trailing edge angle characteristic, deviates from the design reference. As a geometric quantity, the trailing edge angle is directly extracted from the measured data of the blade cross-sectional profile. Its measured value is independent of common-mode interference such as sensor electronic signal drift and ambient temperature changes, thus eliminating the influence of common-mode interference on the trailing edge angle characteristic from the data source. Based on the method of averaging multiple measurements within the monitoring period, random errors caused by instantaneous wind fluctuations or measurement noise in a single measurement are further eliminated. This allows the trailing edge angle characteristic to stably reflect the actual deviation of the geometry at the trailing edge position along the spanwise direction, thereby improving the stability of the trailing edge angle characteristic in response to changes in the blade cross-sectional geometry and the reliability of risk tendency judgment.

[0041] Specifically, the risk analysis module also includes, The instability characteristic calculation subunit is used to determine the mean of the peak time difference within the monitoring period as the cyclone instability characteristic; wherein, the peak time difference is determined based on the time interval between the peak times at adjacent measuring points, and the peak time is the moment when the amplitude of the vortex shedding signal exceeds the reference amplitude.

[0042] It is understandable that the amplitude of the vortex shedding signal is the sound pressure intensity generated by the airflow disturbance at the trailing edge of the blade. The sound pressure signal measured by the acoustic acquisition unit can be represented as voltage or digital quantity after analog-to-digital conversion. The magnitude of the amplitude reflects the strength of the airflow disturbance, and the moment when the amplitude exceeds the reference amplitude is marked as the peak moment.

[0043] In this embodiment, the reference amplitude is the boundary between the periodic sound pressure fluctuations generated by vortex shedding of the blade under healthy operating conditions and the background noise. As an example, those skilled in the art can select the healthy operating period of the blade during the initial operation of the wind turbine, collect vortex shedding signals at each spanwise position of the trailing edge of the blade, and extract the average value of the noise signal amplitude during this period as the background noise reference. The specific value of this reference does not constitute a limitation of the present invention and will not be elaborated here.

[0044] In actual wind turbine operation scenarios, the spanwise non-uniform distribution of material loss on the upper and lower surfaces directly disrupts the originally orderly alternating shedding process of the trailing vortex street. Once the arcuate relationship at the trailing edge of the upper and lower surfaces becomes unbalanced due to material loss, the timing and mode of airflow separation on the upper and lower sides no longer coincide, the vortex shedding rhythm becomes disordered, and the time interval between noise peaks at adjacent spanwise positions becomes scattered and uncertain. Consequently, the cyclone instability characteristics based on the statistical mean of the time differences between each peak increase. The peak time difference eliminates the influence of common-mode interference on cyclone instability characteristics from the data source, enabling the cyclone instability characteristics to stably reflect the actual degree of deterioration of the synchronicity of vortex street shedding at that spanwise position, thereby improving the early warning system's ability to perceive early structural degradation and the accuracy of early warnings.

[0045] Specifically, the spanwise interval corresponding to the measuring point contains several spanwise positions.

[0046] In this embodiment, the vortex shedding signal can be acquired by an acoustic sensor fixed to the nacelle shell or by a non-contact acoustic acquisition device deployed at the bottom of the tower. The vortex shedding component has a narrow-band characteristic synchronized with the blade rotation frequency. It is separated from the acquired sound pressure signal by bandpass filtering to obtain the vortex shedding signal. Bandpass filtering and other signal processing methods are conventional techniques in the field. As long as they can achieve the separation of vortex shedding signals in different spanwise sections, they will not be elaborated further.

[0047] It should be noted that the spanwise interval corresponding to the measurement point contains several spanwise position points, and the correspondence between each measurement point and these spanwise position points is determined based on a frequency-spanwise mapping table. This frequency-spanwise mapping table is established when the blade is in a healthy state. As an example, those skilled in the art can use a laser vibrometer to scan the trailing edge region along the blade spanwise position during the blade's factory inspection phase or during the first shutdown maintenance after the wind turbine is put into operation, obtaining the surface vibration spectrum at each spanwise position, extracting the vortex shedding characteristic frequency, and establishing the correspondence between each spanwise interval and the vortex shedding characteristic frequency. Alternatively, the frequency-spanwise mapping table can be determined theoretically based on the known geometric parameters of each position along the blade spanwise from the hub center and the rated rotational speed. In actual monitoring, a bandpass filter bank is used to separate the acquired vortex shedding signal into narrowband signals corresponding to each spanwise interval according to the characteristic frequencies corresponding to each spanwise interval in the mapping table. Each narrowband signal serves as a measurement point, and adjacent narrowbands are adjacent measurement points.

[0048] Understandably, spanwise frequency mapping tables can be applied throughout the blade's entire lifespan. When asymmetric degradation occurs at the blade's trailing edge, the vortex shedding frequency in the corresponding spanwise region may shift. However, the signal in this region still primarily falls within the filter channel specified in the mapping table. The frequency shift itself does not change the correspondence between the spanwise region and the filter channel; instead, it manifests as an increase in the peak time difference of the vortex shedding signal within this region, which is then captured by cyclone instability characteristics. Bandpass filtering and mapping table creation are conventional techniques in this field and will not be elaborated upon further.

[0049] Specifically, the propagation analysis module includes, The spread range delineation sub-unit is used to define the spread distribution range as the longitudinal area between the spread start position and the spread end position.

[0050] Specifically, the propagation begins at the point where the trailing edge angle feature first decreases unidirectionally along the span, and ends at the point where the trailing edge angle feature increases in the opposite direction along the span.

[0051] In this embodiment, the risk analysis module has identified several suspicious sub-regions with potential risks. Each suspicious sub-region contains multiple spanwise locations, and each spanwise location corresponds to the average monitoring period of a trailing edge angle feature. Starting from the leaf tip and moving towards the leaf root, the trailing edge angle features of adjacent spanwise locations are compared sequentially. The spanwise location where the trailing edge angle feature first shows a unidirectional decrease along the spanwise direction is determined as the propagation initiation location. When the trailing edge angle feature shows a reverse increase along the spanwise direction, it indicates that the location has reached the degradation boundary, and this location is determined as the propagation termination location. The spanwise segment between the propagation initiation location and the propagation termination location is the propagation distribution range. The number of locations used to determine unidirectional decrease and reverse increase can be set according to the spanwise resolution and monitoring accuracy requirements, which will not be elaborated further here.

[0052] Understandably, regarding the extent of the spread, early warning information can be generated by providing information on the spanwise segments corresponding to the spread and the deviation of the trailing edge angle. This information can then be pushed out through the wind farm monitoring system or through a portable detection terminal. Only a one-time early warning push is needed for the spanwise segments between the start and end points of the spread; further details are omitted here.

[0053] In actual wind turbine operation scenarios, once trailing edge material loss occurs in a localized section, the continuous effects of wind and rain erosion and aerodynamic loads drive the degradation boundary to extend longitudinally. This embodiment utilizes the changing trend of the trailing edge angle characteristics to track the degradation front. As the transition from the degraded zone to the undegraded zone occurs, the deviation of the cross-sectional geometric angle decreases, and the trailing edge angle characteristic decreases unidirectionally along the longitudinal direction. The starting point of this trend is determined as the propagation initiation location. When approaching the boundary between the degradation boundary and healthy material, the trailing edge angle characteristic increases in the opposite direction; this location is determined as the propagation termination location. This delineation of the propagation distribution range eliminates the fragmented warning range problem caused by risk tendency determination only identifying discrete risk points and failing to determine the actual extent of degradation. Based on this, the warning push module provides a one-time warning for the entire section, providing a clear repair boundary for operation and maintenance decisions, thereby improving the accuracy of the warning range and the actual damaged section of the blade.

[0054] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A wind turbine fault early warning system based on big data, characterized in that, include, The spanwise segmentation module is used to determine the segmentation position of each blade based on the spanwise mistuning degree; wherein, the spanwise mistuning degree is determined based on the arc length of the upper surface and the arc length of the lower surface at the same spanwise position of the blade. The region selection module is used to determine suspicious sub-regions based on the positional relationship of the segmentation points; The risk analysis module is used to determine the risk tendency of suspicious sub-regions based on trailing edge angle characteristics and cyclone instability characteristics. The trailing edge angle characteristics are determined based on the angle between the trailing edge tangent of the upper arc surface and the trailing edge tangent of the lower surface at the same spanwise position. The cyclone instability characteristics are determined based on the peak time difference of the vortex shedding signal at adjacent spanwise positions. The spread analysis module is used to determine the spread distribution range based on the changing trend of the trailing edge angle characteristics of suspicious sub-regions with risk potential; The early warning push module is used to push early warnings based on the spread and distribution range.

2. The wind turbine fault early warning system based on big data according to claim 1, characterized in that, The spanwise segmentation module includes, The spanwise mistuning analysis sub-unit is used to determine the spanwise mistuning degree based on the ratio of the upper surface arc length to the lower surface arc length at the same spanwise position of the blade.

3. The wind turbine fault early warning system based on big data according to claim 1, characterized in that, The spanwise segmentation module also includes, The segmentation location determination sub-element is used to determine the segmentation location point where the difference between the spanwise mistuning degree and the reference mistuning degree exceeds the reference deviation value.

4. The wind turbine fault early warning system based on big data according to claim 1, characterized in that, The region selection module includes, Sub-regions are defined as sub-units, which are used to identify the blade region enclosed by the upper and lower surfaces between adjacent segmentation points as a suspicious sub-region.

5. The wind turbine fault early warning system based on big data according to claim 1, characterized in that, The risk analysis module includes, The risk tendency determination subunit is used to determine that the suspected sub-region has a risk tendency when the trailing edge angle feature of the suspected sub-region is lower than the reference trailing edge angle value and the cyclone instability feature exceeds the reference cyclone instability value.

6. The wind turbine fault early warning system based on big data according to claim 1, characterized in that, The risk analysis module also includes, The included angle feature calculation subunit is used to determine the mean value of the included angle between the trailing edge tangent of the upper arc surface and the trailing edge tangent of the lower surface at the same spanwise position within the monitoring period as the included angle feature.

7. The wind turbine fault early warning system based on big data according to claim 1, characterized in that, The risk analysis module also includes, The instability characteristic calculation subunit is used to determine the mean of the peak time difference within the monitoring period as the cyclone instability characteristic; wherein, the peak time difference is determined based on the time interval between the peak times at adjacent measuring points, and the peak time is the moment when the amplitude of the vortex shedding signal exceeds the reference amplitude.

8. The wind turbine fault early warning system based on big data according to claim 7, characterized in that, The spanwise interval corresponding to the measuring point contains several spanwise positions.

9. The wind turbine fault early warning system based on big data according to claim 1, characterized in that, The propagation analysis module includes, The spread range delineation sub-unit is used to define the spread distribution range as the longitudinal area between the spread start position and the spread end position.

10. The wind turbine fault early warning system based on big data according to claim 9, characterized in that, The propagation initiation position is the position where the trailing edge angle feature first shows a unidirectional decrease along the span, and the propagation termination position is the position where the trailing edge angle feature shows a reverse increase along the span.

Images