Yaw drive monitoring method

Abstract

A method of monitoring a yaw system (2) of a wind turbine comprising a plurality of yaw drives (D1,..., D8) is described, the method comprising for each yaw drive (D1): receiving encoder data (Eout) from a motor speed encoder (E) of the yaw drive (D1); calculating a signal mean (10M) based on the received encoder data (Eout); calculating a collective signal mean (10Mcoll) from the signal means (10M) of the other yaw drives (D2,..., D8); comparing the signal mean (10M) with the collective signal mean (10Mcoll); and flagging the yaw drive (D1) as potentially faulty if the respective comparison result (10Δ) exceeds a predetermined threshold (10T). A monitoring device (1) for a yaw system (20) of a wind turbine is also described; as well as a wind turbine (2) comprising such a monitoring device (1).

Description

Background Technology

[0001] To keep the wind turbine hub facing the windward direction, the entire nacelle yaws around the top of the wind turbine tower. In widely used configurations, a ring gear (or "gear ring") is mounted at the top of the tower, and a device for controlling a pinion, typically mounted to the nacelle base, is used to rotate the nacelle relative to the fixed tower. Typically, the yaw system implements multiple essentially identical yaw drives, for example, two sets of four yaw drives, one set on each side of the nacelle base. A yaw drive typically includes an electric motor, a gearbox, and a pinion that converts a high motor speed to a lower speed, and teeth that engage with the ring gear. The yaw drive may be equipped with a motor speed encoder (typically an incremental encoder or the like) configured to report motor shaft position data, for example, using two orthogonal output channels, each reporting a fixed number of pulses per complete rotation of the motor shaft. The encoder data from the yaw drive can be reported to the wind turbine controller. Based on the encoder's pulse frequency, the wind turbine controller can then determine the rotational speed of the corresponding yaw drive motor, for example, during yaw action.

[0002] The input-output ratio of a yaw drive gearbox can be very high to generate the required torque. The strong forces acting on the gearbox can cause damage due to wear and tear. In the worst case, the sun gear of the planetary gearbox stage may break, causing the pinion to disconnect from the motor drive shaft. In this failure scenario, the yaw drive motor may continue to operate normally, and the encoder will continue to report a seemingly normal (but slightly higher) motor speed. However, the torque for the ring gear must be provided by other yaw drives. Typically, this type of failure in the gearbox will only be detected during the annual overhaul of the wind turbine. However, if the yaw drive gearbox fails shortly after the overhaul, the remaining yaw drives must share the load until the next annual overhaul. Therefore, wear and tear on these yaw drives is higher than usual, leading to an increased failure rate. Summary of the Invention

[0003] Therefore, the object of the present invention is to provide an improved method for detecting faults in the yaw drive gearbox.

[0004] This objective is achieved by the claimed method for monitoring the yaw drive of a wind turbine yaw system and by the claimed yaw drive monitoring device.

[0005] In the context of this invention, a wind turbine yaw system should be understood to include a gear ring (mounted on the upper part of the wind turbine tower) and multiple yaw actuators, each yaw actuator including an electric motor, a gearbox, and a pinion. The gear ring has teeth around a vertical surface, which can be the outer or inner surface of the gear ring. The yaw actuators can be arranged equidistantly around the gear ring, in different groups, or in any suitable configuration, and are positioned such that the pinion engages with the gear ring. The yaw actuators collectively deliver sufficient torque to rotate the nacelle about the tower axis. The direction of rotation of the motors determines the direction of rotation of the nacelle. Typically, the wind turbine controller initiates yaw action by sending appropriate signals to the yaw actuator motors, causing these yaw actuator motors to be actuated collectively to drive the nacelle in the desired direction. Each yaw actuator should be understood to include a motor speed encoder, which is, for example, arranged in a bearing between the motor and the gearbox. The speed encoder generates digital output data sampled at a certain sampling rate.

[0006] According to the present invention, a method for monitoring a wind turbine yaw system having multiple yaw drives includes: for each individual yaw drive, receiving encoder data from the motor speed encoder of that yaw drive; and calculating an arithmetic mean or "signal mean" based on the received encoder data. The "signal mean" should be understood as representing a single value of the encoder data during the yaw action. A ensemble signal mean is also calculated based on the signal means of the other yaw drives. The signal mean of each individual yaw drive is compared with the corresponding ensemble signal mean, and if the comparison result exceeds a predetermined threshold, the yaw drive is marked as "potentially faulty".

[0007] This invention is based on the insight that the task of rotating the nacelle is more or less equally distributed among the yaw actuators, such that in a fault-free yaw system with all “healthy” yaw actuators, the results of various comparisons should be very similar. As the inventors have noted, if the yaw actuator gearbox is damaged to the point that the motor drive shaft is no longer connected to the pinion, the data reported by the encoder of that yaw actuator will differ from the encoder data reported by the “working” yaw actuator. Therefore, any comparison that significantly deviates from the expected results (a comparison between the arithmetic mean of the arithmetic signal of a particular yaw actuator and the aggregate arithmetic mean representing the other yaw actuators) is interpreted by the method of this invention as a potential fault in the gearbox of the corresponding yaw actuator.

[0008] The advantage of this invention lies in its novel use of readily available information. It is well known to equip electric motors with speed sensors (typically incremental encoders or the like) positioned to measure the rotational speed of the yaw motor. When deployed in the yaw drive of a yaw system, such an encoder reports motor shaft position data to a wind turbine controller, which can evaluate the data to determine, for example, the motor speed of the yaw drive during a yaw braking test procedure.

[0009] According to the present invention, a monitoring device for a wind turbine yaw system includes substantially the same computing modules for each yaw actuator, and each computing module includes: an input stage for receiving encoder data from the motor speed encoder of the respective yaw actuator; a processing stage configured to calculate a signal mean based on the received encoder data; a processing stage configured to calculate a ensemble signal mean based on the signal mean of other yaw actuators; a comparator stage configured to compare the signal mean with the ensemble signal mean; and an output stage configured to mark the yaw actuator as potentially faulty when the corresponding comparison result exceeds a predetermined threshold.

[0010] The monitoring device can utilize the encoder already present in the yaw drive of the yaw system; that is, if each yaw drive is already equipped with a motor speed encoder, no additional hardware is required. The computational stage of the monitoring device of the present invention can be provided as a computer program that can be directly loaded into the memory of the wind turbine controller, the computer program having program elements for executing the steps of the method of the present invention when the computer program is executed by the wind turbine controller. Therefore, the monitoring device of the present invention can be implemented at an advantageous low cost, since it is sufficient to update the wind turbine controller with relevant software elements to execute the present invention.

[0011] According to the present invention, a wind turbine includes an example of a yaw system and the monitoring device of the present invention, wherein the yaw system includes a gear ring and a plurality of yaw actuators, wherein each yaw actuator includes an electric motor, a gearbox and a pinion arranged to engage with the gear ring.

[0012] Particularly advantageous embodiments and features of the invention are given by the dependent claims, as disclosed in the following description. Features from different categories of claims may be suitably combined to give further embodiments not described herein.

[0013] Any suitable motor speed encoder can be implemented in the yaw drive. In the following, without limiting the invention in any way, it can be assumed that the motor speed encoder is an incremental rotary encoder or shaft encoder as indicated above, configured to report position information using two orthogonal output channels. Each output channel “reports” a pulse train (a transition between 0 and 1), with a known number of pulses per revolution. The number of pulses per revolution is the encoder resolution. As those skilled in the art will know, pulse counting or pulse timing techniques can be implemented to determine the pulse frequency, from which the encoder's rotational speed (and therefore the yaw drive motor's speed) can be inferred. The encoder's rotational direction is inferred from the pilot channel.

[0014] As will become clear below, the method of the present invention does not require knowledge of the actual motor speed (rpm and direction of rotation), but rather infers potential faults in the yaw drive gearbox by comparing other quantities derived from encoder data.

[0015] In one approach, encoder data can be processed using DSP techniques such as frequency domain analysis and spectrum analysis. For example, encoder data for each yaw drive can be processed to convert time-domain samples into a frequency-domain representation; the coefficients obtained in this way for each yaw drive encoder can be averaged or otherwise processed to obtain a signal mean, which can be compared with each other to detect potentially faulty yaw drives. However, this technique can involve significant computational effort. In a particularly preferred embodiment of the invention, encoder data is processed to obtain a pulse frequency signal, which is then subjected to zero-mean normalization, i.e., to obtain a signal oscillating around zero. A suitable arithmetic mean of the zero-mean normalized signal is then calculated. In a particularly preferred embodiment of the invention, this step of obtaining the signal mean includes calculating the root mean square (rms) of the zero-mean normalized signal.

[0016] For each sampling moment, the values ​​of the pulse frequency signal, the zero-mean normalized signal, the signal mean, etc., should be understood as digital values. This sequence of digital values ​​is referred to herein as a “curve” or “signal.” The step of processing encoder data to obtain the signal mean is performed at appropriate intervals, such as over a complete yaw maneuver, a portion of a yaw maneuver, or at a predetermined minimum time interval. A yaw maneuver can be initiated whenever the wind turbine controller determines a yaw error, i.e., when the hub is not directly facing the windward direction. To reduce the yaw error to an acceptable level, the yaw actuator is actuated to move the nacelle so that the hub is again facing the windward direction. In a particularly preferred embodiment of the invention, the pulse frequency signal is zero-mean normalized by calculating the root mean square (RMS) of the pulse frequency signal and subtracting the RMS signal from the pulse frequency signal. Zero-mean normalized signals are obtained from the encoder data of each yaw drive, such that in an exemplary configuration with eight yaw drives, the monitoring device calculates eight such zero-mean normalized signals, and the mean of each set can be simply obtained by summing the seven zero-mean normalized signals and dividing by seven.

[0017] The rate at which yaw actions are performed can depend on various factors. Any change in wind direction may require corrective yaw actions. Wind direction can change frequently, depending on weather conditions. During a yaw action, the encoder's pulse frequency signal begins with an upward ramp, followed by a segment where the pulse frequency exhibits essentially oscillatory behavior near a higher level. To obtain a zero-mean normalized signal for the yaw driver, encoder data associated with the initial ramp is preferably discarded, and signal processing is preferably performed on a portion of the subsequent oscillatory segment.

[0018] In a particularly preferred embodiment of the invention, the monitoring device maintains a fault matrix or fault table, which records the cumulative "potential fault" flags for each yaw drive. This table is initialized to empty (after commissioning, after annual inspection, etc.) and then gradually populated by the monitoring device as encoder data is evaluated. In one exemplary embodiment, the fault matrix can be visualized as a two-dimensional table with a "yaw action" column and columns for each yaw drive. In each row, a yaw action timestamp and / or cumulative value can be updated, subsequently resulting in a "0" entry for each healthy yaw drive and a "1" entry for each potentially faulty yaw drive. The table can include a fixed number of rows (e.g., fifteen, twenty, etc.) and can be maintained on a "first-in, first-out" basis.

[0019] Alternatively or additionally, the monitoring device may maintain a fault count for each yaw drive. The fault count may be incremented whenever the yaw drive is marked as “potentially faulty.” In a particularly preferred embodiment of the invention, the monitoring device may be configured to cause the wind turbine controller to issue a fault report (e.g., to the wind farm operator) when the yaw drive is marked as “potentially faulty” after a predetermined number of yaw actions.

[0020] As indicated above, the rotary encoder delivers data that can be used to determine the rotational speed of the yaw drive motor, which may be relevant during yaw braking tests. For the method of the present invention to identify gearbox faults, the actual motor speed is not actually necessary: ​​only the difference between the individual signal mean and the average signal mean is of interest. Data reported by a faulty rotary encoder may therefore be misinterpreted as indicating a faulty yaw drive gearbox. Therefore, the method of the present invention preferably also includes the step of analyzing the encoder output to determine the operating state of the encoder. In a particularly preferred embodiment of the invention, the method includes the step of counting pulses received from the encoder. When the encoder is functioning normally and the connection between the encoder and the controller is fault-free, the pulse count will vary with the motor speed. Similarly, a pulse count remaining at zero may indicate that the encoder is not rotating, and a zero pulse count can therefore be interpreted as indicating a defective encoder. The monitoring device can cause the wind turbine controller to report the fault and schedule a replacement procedure, and exclude the corresponding yaw drive from the monitoring method of the present invention until the encoder is replaced. Attached Figure Description

[0021] Other objects and features of the invention will become apparent from the following detailed description taken in conjunction with the accompanying drawings. However, it should be understood that the drawings are intended for illustrative purposes only and are not intended to define limitations on the invention.

[0022] Figure 1 This is a schematic diagram of a commonly used wind turbine yaw system; Figure 2 This is a block diagram illustrating the monitoring device of the present invention; Figure 3 Exemplary signals for a series of yaw operations are shown when performing the method of the present invention; Figure 4 This is a schematic diagram of a yaw drive fault record generated during the execution of the method of the present invention; Figure 5 This is a schematic diagram of the monitoring device 1 of the present invention.

[0023] In the accompanying drawings, the same numbers always refer to the same objects. The objects in the accompanying drawings are not necessarily drawn to scale. Detailed Implementation

[0024] Figure 1This is a schematic diagram of a commonly used wind turbine yaw system 20. The diagram shows a nacelle 24 with aerodynamic rotors 25, 26 mounted on top of a tower 23. These rotors should always face the windward direction to maximize power output and minimize rotor blade load. For this purpose, the wind turbine 2 is equipped with a yaw system 20 having several yaw drives, in this case eight yaw drives D1-D8, arranged to rotate the nacelle 24 about the vertical axis of the tower 23. Each yaw drive has a motor 210, a gearbox 211 (typically a planetary gearbox), and a pinion 212, as shown in the enlarged section. The pinion 212 engages with a gear ring 22 at the top of the tower 23. The wind turbine controller 29 can initiate yaw action, for example by issuing an appropriate command 290, causing the yaw drive motors 210 to collectively drive the nacelle 240 in the desired direction. Typically, yaw action involves driving the nacelle until the yaw error is reduced to an acceptable value. Hub alignment can be secured by applying a yaw brake. For wind turbine controller 29, it is important to ensure that the yaw brake is functioning properly. For this purpose, rotary encoders E are typically arranged in the yaw drives D1-D8 to report the motor shaft position. As will be familiar to those skilled in the art, rotary encoders E output pulse sequences on two channels, and the motor speed and direction of rotation can be inferred using appropriate techniques such as pulse counting or pulse timing.

[0025] Ideally, the yaw operation of the nacelle 24 is evenly distributed among the yaw drives D1-D8. However, gearbox 211 may fail for various reasons. In this case, the motor 210 of the failed yaw drive D1-D8 will no longer rotate the pinion 212, and the other yaw drives must perform all the work until the defect is discovered. Typically, this failure is only detected during routine maintenance procedures, which may only occur rarely, particularly in the case of offshore wind turbines.

[0026] Figure 2 This is a schematic block diagram illustrating the monitoring device 1 of the present invention. The diagram shows a calculation module 10 for processing data generated by the motor speed encoder of one of the multiple yaw drives D1-D8 of the wind turbine yaw system 2, and it should be understood that the same calculation module 10 is provided for each yaw drive. Here, the processing steps are shown for one of the eight yaw drives D1-D8 in the group. During yaw operation, the received encoder data Eout is processed in the input stage 101 to obtain a pulse frequency signal 10PF. In this exemplary embodiment, a zero-mean calculation stage 102 calculates a zero-mean normalized signal 10ZM based on the pulse frequency signal 10PF, and a mean calculation stage 103 calculates the signal mean 10M, such as the root mean square 10M, of the zero-mean normalized signal 10ZM.

[0027] The calculation module 10 also receives signal averages 10M from the calculation modules 10 of the other seven yaw drives D2–D8. These seven signal averages 10M are averaged in block 104 to obtain a ensemble average 10Mcoll for the remaining yaw drives D2–D8. The ensemble average 10Mcoll and the local average 10M are then compared in block 105. For example, the ensemble average 10Mcoll can be divided by the local average 10M and multiplied by 100 to obtain a percentage value 10Δ, quantifying the degree to which the encoder output from this yaw drive differs from the encoder outputs from the other yaw drives. The result 10Δ is compared with an allowable threshold 10T in block 106. In a healthy yaw system, the local average 10M is expected to be very similar to the ensemble average 10Mcoll. However, if yaw drive D1 fails, the local average 10M will differ significantly from the ensemble average 10Mcoll. A suitable threshold 10T (e.g., a percentage of the ensemble mean 10Mcoll) is applied by comparator 106 to identify any such deviation. Whenever the local mean 10M differs from the ensemble mean 10Mcoll by more than the acceptable threshold 10T, the gearbox of yaw drive D1 can be marked as "potentially faulty." For example, the gearbox state DS1 of yaw drive D1 can then allow the fault matrix to be updated by incrementing the fault count of yaw drive D1.

[0028] This diagram illustrates an optional add-on to monitoring device 1, in the form of an encoder monitoring block 108, which performs pulse counting on encoder data Eout to identify faulty encoders. In the event of an encoder failure in yaw drive D1, the encoder will not generate position data even if yaw drive D1 is otherwise operating normally. Due to the "zero speed" reported by the faulty encoder, the local mean 10M will be practically zero, which will be significantly different from the ensemble mean 10Mcoll. However, by also considering the corresponding encoder state 10E (e.g., a flag that remains "0" as long as the encoder is operating normally and can be set to "1" to indicate that the encoder is no longer delivering useful data), comparator 106 will not flag the yaw drive as potentially faulty. Instead, this information can be used to issue a fault report, allowing the wind turbine operator to schedule maintenance procedures for the yaw drive D1. Encoder state 10E can be reported as output signals ES1, ..., ES8 to the fault monitoring level, as will be explained below.

[0029] Figure 3 As shown above Figure 2The yaw system with eight yaw drives D1–D8 described herein executes the method of the present invention with an exemplary zero-mean normalized signal 10ZM of a series of yaw operations Y1, Y2, Y3. Each yaw operation Y1, Y2, Y3 can last for several minutes until the hub is correctly aligned with the windward direction and the yaw error has been reduced to an acceptable level.

[0030] In the lower part of the graph, the chart is magnified between time T1 and time T2 to show a portion of the yaw action Y2 in more detail. Each zero-mean normalized signal 10 oz is as described above. Figure 2 As described in the figure, seven of the eight 10ZM signals are very similar, such that when any one of these seven (corresponding to...) is... Figure 2 When comparing the mean 10M of the yaw drives (D2–D8) with the ensemble mean 10Mcoll, similar results can be expected. However, when the remaining signal mean 10M (corresponding to Figure 2 When comparing the yaw drive D1 in the dataset with the set mean 10Mcoll, the comparison results will be significantly different. Significantly different comparison results can be interpreted as an indication that the corresponding yaw drive D1 has a faulty gearbox. For example... Figure 2 As explained in the document, the corresponding encoder status 10E can be evaluated to ensure that the fault does not lie in the encoder itself.

[0031] By calculating and comparing the mean values ​​of different types of signals as described above, the dynamic characteristics of the signal can be extracted from the encoder data. Therefore, even if the motor speed of a "faulty gearbox" yaw drive appears similar to that of a healthy yaw drive, this invention allows for the confident identification of a "faulty gearbox" yaw drive. This information can be used to prevent overloading of the healthy yaw drive.

[0032] Figure 4 The yaw drive fault matrix M is generated when the method of this invention is executed. DS A schematic diagram of (or "fault log"). Fault Matrix M DS The data is arranged in a table, with the left column used for an incrementing count of yaw actions, and one column for each of the eight yaw drives. The status signals D1, ..., D8 of each yaw drive are recorded in the table – for example, "1" indicates "faulty" and "0" indicates "non-faulty". Fault record M DSThe system can store the results of the most recent n yaw actions, such as 20 as shown here, appropriately overwriting older entries. Here, the column for yaw drive #5 indicates a persistent fault in its gearbox, and the wind turbine controller can issue a fault report when the fault count reaches a predetermined limit. For example, if ten or more faults are recorded for yaw drive #5 in the most recent 20 yaw actions, the wind turbine controller can issue a fault report for yaw drive #5. This information can be used by the wind turbine operator when preparing for annual inspections or to schedule additional maintenance procedures. Minor faults (such as those indicated in the column for yaw drive #3) are recorded but do not necessarily result in a fault report.

[0033] A similar table (not shown here) can be used to record encoder health states ES1, ..., ES8, for example, a table with a left column representing yaw actions and a column for each yaw drive. For each yaw action, the encoder states ES1, ..., ES8 for each yaw drive D1, ..., D8 are recorded. Whenever monitoring device 1 identifies a faulty encoder (e.g., the corresponding signal ES1, ..., ES8 is set to "1"), the corresponding entry is recorded in the table. Similar to the yaw drive fault matrix described above, this encoder state table can store the results of the most recent n yaw actions, appropriately overwriting older entries. Also here, the count of fault state flags can lead to fault reports, for example, if the encoder states ES1, ..., ES8 of the yaw drive have been set to "1" a certain number of times in the most recent 20 yaw actions, and this information can be used by the wind turbine operator when preparing for annual inspections or to schedule additional maintenance procedures.

[0034] Figure 5 This is a schematic diagram of the monitoring device 1 of the present invention. The diagram illustrates how encoder data from each of the multiple yaw drives D1–D8 is collected and processed in the calculation module 10 as described above to generate gearbox status values ​​DS1–DS8 for each yaw drive D1–D8, for example, after a completed yaw action. The gearbox status values ​​DS1–DS8 are evaluated in the fault assessment module 15, which appropriately updates the fault matrix M. DS And a fault report 150 is issued whenever necessary. Similarly, encoder state values ​​ES1–ES8 are evaluated in fault assessment module 15, which updates the state matrix M appropriately. ES And a fault report 150 is issued whenever necessary. For example, a fault report 150 may be issued if the fault count of the yaw drives D1–D8 has reached or exceeded the limit during a predetermined number of yaw actions; a fault report 150 may be issued if the encoder E of the yaw drives D1–D8 is deemed to be faulty.

[0035] Although the invention has been disclosed in the form of preferred embodiments and variations thereof, it should be understood that many other modifications and variations may be made thereto without departing from the scope of the invention. For example, although the invention has been described using an example of a yaw system with eight yaw drives, it should be understood that the yaw system may include any number of yaw drives.

[0036] For clarity, it should be understood that the use of “a” or “an” throughout this application does not exclude a plurality, and “including” does not exclude other steps or elements.

Claims

1. A method for monitoring a wind turbine yaw system (2) comprising multiple yaw actuators (D1, ..., D8), the method comprising, for each yaw actuator (D1): - Receive encoder data (Eout) from the motor speed encoder (E) of the yaw drive (D1). - Calculate the signal mean (10M) based on the received encoder data (Eout). - Calculate the ensemble signal mean (10Mcoll) based on the signal mean (10M) of the other yaw drives (D2, ..., D8). - Compare the signal mean (10M) with the ensemble signal mean (10Mcoll); and - If the corresponding comparison result (10Δ) exceeds the predetermined threshold (10T), the yaw drive (D1) is marked as potentially faulty.

2. The method according to the preceding claim includes the steps of processing encoder data (Eout) to obtain a pulse frequency signal (10PF) and performing zero-mean normalization on the pulse frequency signal (10PF).

3. The method according to the preceding claim, wherein, The steps for calculating the signal mean (10M) include calculating the root mean square of the zero-mean normalized signal (10ZM).

4. The method according to any one of the preceding claims, wherein, The encoder signal (Eout) is collected during a portion of the yaw action (Y1, Y2, Y3).

5. The method according to any one of the preceding claims, further comprising the following step: recording the fault status (DS1, ..., DS8) of each yaw drive (D1, ..., D8) in a fault log (M DS )middle.

6. The method according to any one of the preceding claims, comprising the step of maintaining a fault count for each yaw drive (D1, ..., D8).

7. The method according to any one of the preceding claims, comprising the step of issuing a fault report (150) when the fault count of the yaw drive (D1, ..., D8) exceeds a predetermined limit.

8. The method according to any one of the preceding claims, comprising the step of determining the operating state (10E; ES1, ..., ES8) of each encoder (E).

9. The method according to the preceding claim, wherein, If the encoder (E) of a potentially faulty yaw drive (D1, ..., D8) is considered to be operating normally, then the yaw drive is marked as faulty.

10. The method according to any one of the preceding claims, wherein, Potentially faulty yaw drives (D1, ..., D8) were excluded from subsequent fault assessments.

11. A monitoring device (1) for a wind turbine yaw system (20) comprising multiple yaw actuators (D1, ..., D8), the monitoring device (1) comprising a computing module (10) for each yaw actuator (D1, ..., D8), wherein, Each computing module (10) includes: - Input stage (101) for receiving encoder data (Eout) from the motor speed encoder (E) of the corresponding yaw drive (D1). - Processing stages (102, 103) are configured to calculate the signal mean (10M) based on the received encoder data (Eout). - Processing stage (105) is configured to calculate the aggregate signal mean (10Mcoll) based on the signal mean (10M) of other yaw drives (D2, ..., D8). - Comparison level (104), configured to compare the signal mean (10M) with the ensemble signal mean (10Mcoll); and - Output stage (106) is configured to mark the yaw drive (D1) as potentially faulty when the corresponding comparison result (10Δ) exceeds a predetermined threshold (10T).

12. The monitoring device according to the preceding claim, wherein, Each calculation module (10) is configured to determine the operating state (10E; ES1, ..., ES8) of the corresponding motor speed encoder (E).

13. A wind turbine (2), comprising: - A yaw system (20), the yaw system comprising a gear ring (22) and a plurality of yaw actuators (21), wherein each yaw actuator (21) comprises an electric motor (210), a motor speed encoder (E), a gearbox (211), and a pinion (212) arranged to engage with the gear ring (22); and - The monitoring device (1) according to any one of claims 11 to 12.

14. The wind turbine according to the preceding claim, wherein, The monitoring device (1) is implemented as a module of the wind turbine controller (29).

15. A computer program product comprising a computer program that can be directly loaded into the memory of a wind turbine controller (29), and the computer program comprising program elements for performing the steps of the method according to any one of claims 1 to 10 when the computer program is executed by the wind turbine controller (29).

Images