The present invention relates to a method of operating a wind turbine generator, WTG, having a tower supporting a nacelle supporting a rotor with blades and a drivetrain; the method comprising one or more acts. There is an act of providing a time series window indicative of vibrations. There is an act of transforming the time series using a set of basis functions into a frequency domain with an amplitude and a phase. There is an act of monitoring the phase for a change in the phase and detecting a change. Upon not detecting a change in the phase, normal operation of the WTG is performed. Upon detecting a change in the phase, corrective action is applied to the operation of the WTG.

Background of the Invention

It is of outmost importance to guarantee a dynamic stability of a wind turbine. Vibra- tions are therefore a source of concern. Stall induced blade vibrations are well-known problems in wind turbines. Even primitive wind turbines with fixed blades or stall- regulated wind turbines had blade vibration issues.

Attempts to resolve these issues by blade dampers were developed to cope with the vibrations.

Modern pitch-regulated turbines can also experience conditions that stall certain part of the blades at certain azimuth angles.

As it is known, stall occurs when the angle of attack increases above a critical angle of attack (AoA) after which the lift starts to decrease and the drag to increase. Stall is detrimental for turbine performance, decreasing the power output. It also causes the above-mentioned blade edgewise vibrations.

When irregular rotor inflow conditions and/or turbine operation cause the local angle of attack to vary as the blade sweeps the rotor plane such that the angle of attack ex- ceeds the critical angle of attack, flow separation stall occurs. Irregular rotor inflow can be, but are not limited to, shear, yaw error, terrain slope, veer, extreme turbulence, wake, or a combination thereof.

Turbine operation sometimes requires power boosting which can cause the angle of attack to increase beyond the critical angle of attack.

Patent publication WO 2016/169963 discloses a method of correcting rotor imbalance and comprises measuring vibrations within at least one time window and determining an imbalance factor and an imbalance phase. The values of the parameters in the equa- tion for calculating the correction action are then updated based on the imbalance factor and an imbalance phase. A correction angle for each of the wind turbine blades is calculated using these adjusted parameters. The correction angle is used to aerody- namically balance the rotor.

However, patent publication WO 2016/169963 is directed towards correcting a mass Imbalance based on monitoring and detecting a change in the mass imbalance. Data are from within a predetermined time window, and the control system determines an imbalance factor and an imbalance phase based on the measured data. The control system analyses the imbalance factor and imbalance phase to detect any changes in the mass imbalance, e.g. relative to at least one previously measured set of data. If the changes indicate that the mass imbalance is changed, e.g. reduced, but is still outside the threshold range, then the location and/or weight of the balancing mass is adjusted. It is an objective to improve operation of a wind turbine by improving dynamic stabil ity. It is an objective to reduce stall events. It is an objective to operate a wind turbine as dynamically stable and optimal as possible.

Description of the Invention

An objective is achieved by a method of operating a wind turbine generator, WTG, having a tower supporting a nacelle supporting a rotor with blades and a drivetrain; the method comprising one or more acts. There is an act of providing a time series window indicative of vibrations. There is an act of transforming the time series using a set of basis functions into a frequency domain with an amplitude and a phase. There is an act of monitoring the phase for a change in the phase and detecting a change. Upon not detecting a change in the phase, normal operation of the WTG is performed. Upon detecting a change in the phase, corrective action is applied to the operation of the WTG.

In example, a change in the phase of the drivetrain rotational speed (RPM) observed or monitored at fixed frequency has shown to indicate a transition to stall. It is noted that in reality stall may happen in a transition region, regime or interval. Thus, a change in phase of a few degrees may indicate a transition and require corrective ac- tion to mitigate. A change in phase of tens of degrees may indicate a transition and may require corrective action to mitigate. A person skilled in the art will be able to determine and set thresholds and levels of acceptance of stall according to operational circumstances.

The observed change in phase has shown to be associated with the application of an external force with the same or similar frequency due to the blades vibrating in phase with each other, which most probably is due to that there are stall conditions.

Thus, when the blade sweeps an area of the rotor plane at which stall conditions are present, a portion of the blade stalls and the sudden loss in lift and increase in drag causes the blade to vibrate.

In a way, the vibration can be considered as an impulse (albeit weak) that can be ob- served in the frequency domain as a sudden change in the phase of a preferred fre- quency. It is noted that the time series may be sensory outputs from different sensors measur ing dynamics of the wind turbine.

There may be an act of decomposing the time series using a set of basis functions into an amplitude and a phase representation. The transformation of the time series or win- dow of time series may be from a time domain to a frequency domain. The transfor mation may be a Fourier Transform. The transform may be performed digitally. The Fourier Transform may be a FFT.

The transform may be based on a windowed Fourier type transform. The window, i.e. the period of time of interest, may be selected in the time domain. The window may be determined by using short time Fourier Transform or a kernel. The transform may be any equivalent transform based on windows. A wavelet-type transform may be used where a certain (threshold) value of the scaling parameter may be indicative of a transition to stall.

This time series may be from vibration signals from different locations or parts of the turbine and a person skilled in the art may explore and adjust sensors, including sensor positions, to obtain vibration signals with the sufficient information for changes in the phase to be detected.

The corrective action based on phase change detection may be applied using multiple control strategies that can be implemented to take advantage of this information.

Vibration monitors can be used to ensure turbine safe operation by, in example, pitch ing the blades to reduce stall or reduce the risk of entering stall. This enables wind turbine performance optimization since the corrective action can increase the margin in which it is possible to adjust the blade pitch angle to maximize power whilst ensuring that no significant stall occurs. This scheme may be robust for varying conditions. A further effect is that long term tracking of stall conditions can be used to detect soil ing and/or blade erosion. Thus, collection of changes in the phase enables detection of such mechanical deteriorations of a wind turbine.

In an aspect, detecting a change is probably based on a change of the phase that is larger than 5 degrees. Thus, a corrective action may be applied when the change in phase is larger than 5 degrees. Considered as a small change in phase, this may also trigger an alert phase or further observation phase. In an example, a corrective action may not be activated, but only prepared. A detection of a change may be likely when based on a change of the phase that is larger than 20 degrees. Thus, corrective action may be applied when the change in phase is larger than 20 degrees. Considered as a medium change in phase, this may also trigger an alert phase or further observation phase as for the small change in phase. Degrees or levels of attention or alerts may be introduced according to the level of change.

A detection of a change may be evident when based on a change of the phase that is in the order of 30-40 degrees. Again, a corrective action may be applied.

A person skilled in the art may implement acts when an algorithm is to detect changes as outlined and to select levels or thresholds according to the before disclosed sug- gested ranges or values.

In an aspect, the change in phase may also be detected by the speed or paste at which the change occurs. Thus, a fast change in phase may detect a change in phase. Hence, a relatively small change in the phase, e.g. 5 degrees occurring fast, may be as indica- tive of a real event as a larger change in the phase, e.g. 30 degrees occurring fast or slow.

In an aspect, the time series window has a duration of (N+a)*(l/f), where N is an inte- ger.

a is a real number, in the range of [0.3-0.7], preferably [0.4-0-6] or about 0.5.

f is the drivetrain fixed frequency .

Examples:

The presented formula is surprisingly simple, workable and robust.

An a- value of 0.5 has surprisingly shown to be particularly simple, workable and ro bust. N-value of 0 giving 0.5*(l/f) i.e. 0.5/f may be used.

For WTGs, a time series window of about 3 to 5 seconds may also be a starting point. In an aspect, the time series includes vibration measurements of the rotor drivetrain.

Drivetrain signals have shown to be useable and easily obtainable with the required information, i.e. phase information, and to enable reliable results.

In an aspect, the time series includes vibration measurements from the nacelle. Vibra- tion measurements may be from accelerometers placed in the nacelle or at the tower. A nacelle vibration has the required information and provides reliable results that can be used.

In an aspect, the time series includes vibration measurements from one or more blades. Blade movements and vibrations may be used as outlined. Blade movements and vibrations may be obtained from sensors, such as accelerometers, gauges, or la- sers.

In an aspect, the time series is provided as a function of torque on one or more blades for a constant rotational speed (RPM).

Operation may be at a rated rotational speed (fixed or constant RPM) at which point operation is on a“straight” torque-line to a maximum torque (Tmax)· From the rated rotational speed, the angle of attack is increased to increase the torque. The value of torque on the straight torque-line may be used as an input. The value of torque may be used as input to determine the time series or time series parameters such as the time window.

Thereby, it is achieved that operation can be sustained without stall. In particular, it is possible to operate closer to a maximum torque value. At least with reduced risk of stall. The operational stall margin, which in terms of angle of attack as an example may be about 4 degrees, is then expanded or at least sustained.

In an aspect, the corrective action is chosen amongst one or more of the following actions. There may be a correction of angle of attack of at least one blade. There may be a cor rection in pitch angle.

There may be a correction in torque. The torque may be corrected indirectly by per- forming actions altering torque. There may be a correction in power. The power may be corrected indirectly by performing actions altering power. There may be a correc- tion in rotational speed settings (RPM- settings). The rotational speed (RPM) may be corrected indirectly. An object may be achieved by a device for operating a wind turbine generator, WTG, having a tower supporting a nacelle supporting a rotor with blades and a drivetrain; the system comprising sensors and means adapted to execute the actions disclosed herein. An object may be achieved by a computer program product comprising instructions to cause the device of claim to execute the actions disclosed herein.

Description of the Drawing

The invention is described by example only and with reference to the drawings, whereon:

Fig. 1 illustrates a wind turbine operating at a rotational speed;

Fig. 2 illustrates time series representing rotational speed (RPM);

Fig. 3 illustrates amplitude components of a transformed time series;

Fig. 4 illustrates phase components of a transformed time series;

Fig. 5 illustrates a RPM torque operational diagram of a WTG; and

Fig. 6 shows exemplary time series from the drivetrain and blade dynamics.

Detailed Description of the Invention

Fig. 1 illustrates a wind turbine generator (WTG) 1 having a tower 2 supporting a na- celle 3 supporting a rotor 4 with a blade 5. The rotor 5 is directly coupled to a drivetrain 6 and rotates with a rotational speed in rounds per minutes (RPM) 8.

The WTG may have (not shown) one or more sensors measuring the rotational speed (RPM) of the rotor. The sensors generate time series 10 that sample at appropriate sampling rates. The time series may be stored within the WTG or transferred to another location such as a central processing facility. The time series may be processed in a computer by appro- priate processing and analytics software. The time series may represent drivetrain sensory outputs. The time series may repre- sent accelerometer sensory outputs. The time series may represent tower movement sensory outputs. The time series may represent blade movement sensory outputs. The time series may represent blade vibration sensory outputs. In the following, the shown time series 10 is an example of a drivetrain 6 sensory out- put.

Figure 2 illustrates a time series 10 obtained as a measure of rotational seed (RPM) 8 of a rotor 4 as exemplified in figure 1 and the time series 10 is obtained from sensory output from a drivetrain 6.

A time series window 12 of the time series 10 is extracted. The time series window 12 shows vibrations 14. The time series window 12 may be a running window that is shifted in time. The time series window 12 may be a window that is detected or estab- lished by identifying minor fluctuations or ripples in the time series 10 data. The time series window 12 is for a constant rotational speed (RPM) 18, which may be a plateau and there may be variations as indicated. Figure 3 illustrates a spectrum of amplitudes 32 as a function of frequency 24 after a transform 30 of the time series 10 from a window 12 as exemplified in figure 2.

Figure 4 illustrates a phase 34 as a function of frequency 24 after a transform 30 of the time series 10 from window 12 as exemplified in figure 2. The shown phase 34 is as- sociated with the amplitude value 32 illustrated in figure 3.

During normal operation, the phase spectrum 35 is monitored and a normal or refer ence phase spectrum 37 (solid line) or a reference phase 37 for a particular frequency is present. The phase spectrum 35 or the phase 34 may change. The phase spectrum 35 (phase as a function of frequency) may be monitored as a whole as a reference phase spectrum 37. The phase spectrum 35 may be monitored for a range or window of fre- quencies. The phase 34 may be monitored for a particular reference phase 37 for a particular frequency such as the drivetrain fixed frequency.

The dotted line shows a change in phase 36 or a deviation from the reference phase 37. In example, the phase 34 is monitored for a particular frequency. Passing a thresh old value detects a change in phase 38. The threshold may be in absolute degrees or a relative value based on the reference phase value.

Figure 5 illustrates a torque 19 vs rotational speed 18 (T vs RPM) operational diagram of a WTG. The WTG will operate at a rated or optimal RPM, here constant RPM 18. The operation is between a T and a T _max at constant rotational speed (RPM 18), and ideally as close to T _max without entering a stall regime.

Figure 6 shows exemplary data over time during normal operation 50 and when cor rective actions 52 are applied. Figure 6 A shows normalised time series 10 from drivetrain 6 sensors showing high speed generator rotational speed (RPM).

Also shown, figure 6B, are alternative data signals (normalised) 9’, 9”, 9’” of edge moment of a blades 5, say 5A, B and C.

Figure 6C shows the pitch position (normalised) of blade 5 A as will be explained as a result of the signal described in figure 6D. Figure 6D shows the phase 34 of the transformed 30 time series 10 window 12 of the high speed generator signal 9 from figure 6A at a frequency (ί=2pw). The phase 34 (at a frequency) is monitored. Changes 36 in the phase are observed. Below a threshold value operation is considered to be normal 50. According to a threshold, setting detection of a change 38, that is a sufficient change 36, is deter mined and here marked by a circle. At this detection of a change 38, corrective action 52 is applied to operation. Figure 6C shows the corrective action 52, which here is a change of angle of attack (AoA) 54 applied as a correction of the pitch angle 55 of blade 5A.