Field of the invention

The present invention relates to a strategy for controlling and regulating the different operational parameter (such as e.g. a blade pitch angle, a position of a flap, and/or other means for changing the aerodynamic surface of a blade) of a wind turbine during operation with a view to reduce the extreme loads acting on the wind turbine tower. The invention furthermore relates to a method for reducing the extreme bending moments at the tower base arising for instance due to wind gusts or turbulence.

Background

During operation, a wind turbine can in general terms be controlled via changing either the torque or the aerodynamic properties of the blades. The latter is achieved by modifying for instance the pitch angle of the blades, the blade length in case of extendable blades, the coning angle or the yaw, flap positions or movements, or other means for changing the aerodynamic surface of the blade such as ailerons, vortex generators etc.

The control activity of modern type wind turbines is mainly governed by the objectives of either optimising the power production or controlling the generator speed. The operation mode of optimising the power production is the governing objective at lower wind speeds where the turbine nominal power has not yet been reached and is denoted 'partial load'. At higher wind speeds a wind turbine is generally controlled with a view to control the generator speed and minimising the loads on the turbine while the power absorbed by the rotor is kept at a constant level. This operation mode is denoted 'full load operation'.

The objectives mentioned above address power production, but only indirectly address the thrust force excerted on the rotor. That is, the nominal operating points inherently ensure that the thrust force is kept within design limits at the stationary operating points. Problems arise when dynamic effects are considered. For instance due to the generator speed-based pitch control being quite a slow-acting system, gusty wind conditions can lead to events where the pitch controller is not able to follow the steep slopes of the wind speed. In effect, the rotor will accelerate away from the stationary operating point, eventually exceeding the overspeed limit and thus initiating an emergency shutdown. Such emergency shutdowns will, in some cases, cause load problems due to negative thrust being applied on the rotor. Excessive loads can, however, also occur during the rotor acceleration leading to the over- speed situation. This is basically caused by the fact that the controller of the operational parameters of the wind turbine such as the pitch controller is either not designed to react on fast changing wind conditions or is not fast enough to compensate for the increased wind speed. Therefore, the thrust coefficient is not lowered fast enough to compensate for the increased wind speed. As a result, the tower is deflected backwards with the result of excessive bending moments in the tower base.

Description of the invention

It is therefore an object of embodiments of the present invention to provide a control system or a control method capable of taking dynamic effects of the wind such as gusts or turbulence into account in the controlling of the operational parameters of wind turbine.

It is a further object of embodiments of the invention to provide a control scheme for a wind turbine reducing the tower vibrations and the extreme tower loads, especially the extreme bending moments near the tower base.

It is a yet further object of embodiments of the invention to provide a control strategy modifying the operational parameters of a wind turbine which is simple and effective and can react fast to changes in the wind while taking the tower deflections into account.

In accordance with the invention this is obtained by a method for controlling and regulating at least one operational parameter of a wind turbine blade (such as e.g. a blade pitch angle, a position of a flap, and/or other means for changing the aerodynamic surface of a blade) on a wind turbine during operation for reducing extreme tower loads, and where the method comprises the steps of:

determining at least one space of acceptable operating situations for the wind turbine based on a set of normal operating situations,

- measuring an acceleration of the nacelle,

determining a velocity of the wind turbine nacelle relative to the ground and based on said acceleration,

determining a position of the wind turbine nacelle relative to a running mean and based on said acceleration, determining the actual operating situation from at least said acceleration, velocity, and relative position,

comparing said actual operating situation to said at least one predetermined space of acceptable operating situations, and based hereon choosing a control strategy from a predetermined set of strategies,

defining a control function for the operational parameter based on said predefined control strategy, and

controlling at least one of the operational parameters of at least one of the wind turbine blades in accordance with said control function.

According to embodiments of the invention, the controlling of at least one operational parameter comprises controlling the pitch angle, a flap position, and/or a mean for modifying the aerodynamic surface of at least one wind turbine blade. The control method could thus further relate to control of ailerons, the activation/deactivation of vortex generators, change of the surface properties or dimensions of the blade etc.

According to the above, the actual operational situation of the wind turbine is characterized by the nacelle acceleration, the velocity of the nacelle, and the position of the wind turbine nacelle relative to a running mean. The latter parameter is throughout the description also called the deviational or relative position or deflection of the nacelle, and describes the deflection of the wind turbine tower relative to the mean deflection in a time interval up to the actual time. The two parameters of the nacelle velocity and the position of the wind turbine nacelle relative to a running mean are determined and derived from the nacelle acceleration. This in turn is measured by means of an accelerometer advantageously placed on the nacelle or in other ways near the top of the tower. The nacelle acceleration can be measured directly as the fore-aft acceleration or can be taken as the size or root mean square value of the ac- celeration vector measured by a two-directional accelerometer. Hereby can be obtained a measure for the operational situation of the wind turbine in a fast and robust way yet being far more accurate and capable of reflecting the operational conditions far more nuanced than other conventional control systems where often only the current wind speed is considered. Another advantage is that the system need only measure the nacelle acceleration which can be done simply and fast, yet precisely. Further, no additional measuring means such as e.g. strain gauges or wind speed anemometers need be used in the above method. The so determined actual operating situation is then according to the present invention compared to a space or confidence volume of operational situations that are acceptable for the wind turbine and for which the operational situation is with a certain minimum probability not extreme nor needs special attention. The proposed control is in this way functioning as an extra control module acting in addition to the normally applied main control system of the wind turbine, but is only active in the extreme and rare operational situations (such as extreme wind gusts or turbulence) outside the aforementioned space or confidence volume of acceptable operating situations. This space is determined advantageously from a large set of normal operation systems for the wind turbine which can be obtained e.g. from simulations of different wind and production conditions, from statistical data, and/or from real-time data collected previously from other wind turbines during normal operation.

A strong correlation is found to be present between the acceptable nacelle accelerations, velocities, and deviational positions, which in one embodiment of the invention leads to describing the space of acceptable operating situations by an ellipsoid in the (acceleration, velocity, deviational position)-space. The control strategy to be employed in reaction to the present conditions is chosen by comparing the actual operating situation to the space or confidence volume of acceptable situations, which in the case of expressing the space as an ellipsoid yields a simple and fast operation.

The predetermined set of control strategies may for instance comprise an activation strategy, a relaxation strategy, and an inactive strategy as will be described in more details in the following. Based on the chosen control strategy is defined a control function expressing how the operational parameter is to be modified and changed, where after the operational parameter is then changed according to said control function.

The acceleration measurements may be performed as often as desirable or even continu- ously. The shorter the time between successive measurements, the more a precise and fast reacting control method can be obtained. Optionally, the nacelle velocity or the relative position need not be updated as often as the nacelle acceleration, which may probably render the proposed control method a little more imprecise and crude in its determination of the actual operating situation, but which on the other hand may increase the speed of the control method allowing the wind turbine to react even more promptly on changes in the wind speed and tower dynamics.

According to a further embodiment of the invention, the determining of the relative position of the wind turbine nacelle is further based on the velocity of the wind turbine nacelle. This is advantageous in resulting in a more accurate and precise yet relatively simple and fast de- termination of the relative position of the tower without the need for other measuring means such as e.g. some kind of position sensors, or strain gauges.

A further embodiment of the invention describes a method according to any of the above, where the normal operating situations are based on statistics, simulations, and/or data from wind turbines operating under normal wind conditions. In this way the normal operating situations for the wind turbine can be pre-determined and analyzed with a view to determine the space of acceptable operating situations. The certainty of the normal operating situations may further be increased as both simulated data, actual operation data and/or statistics may be used together.

According to an embodiment of the invention, the normal operating situations are determined at a maximally allowed wind speed whereby a conservative control method is ensured, in that the used normal operating situations then represent a worst case scenario for the wind turbine. Hereby is obtained, that the control method will only be activated in the rare wind and operating situations leading to extreme tower loads.

According to another embodiment of the invention, the control method comprises determining a set of spaces of acceptable operating situations for the wind turbine, each space being acceptable for a predefined range of wind speeds.

According to an embodiment of the invention, the actual operating situation is further determined from the actual wind speed.

In this way the actual wind speed may also be taken into account in the control method, as the actual operating situation may then be compared to the space of acceptable operating situations for the same wind speed or range of wind speeds. Hereby the control method can react more precisely and refined to the present operating condition. Similarly, the space or confidence volume of acceptable operating situations may be refined to depend also on other wind turbine parameters such as for instance the temperature, humidity, the yaw- misalignment,- all of which are parameters known to the influence operation conditions of a wind turbine.

According to yet another embodiment of the invention, the control strategy is further chosen based on the velocity of the wind turbine nacelle, on the change of the actual operating situa- tion with respect to a change in the nacelle acceleration, and/or on the value of the actual operational parameter. Hereby is obtained a refinement of the control strategy, in that it may be chosen not only based on whether the actual operational situation is acceptable or not (lies within or outside the confidence volume) but also based on considerations on for instance if a certain change in the operational parameter will most probably ameliorate or worsen the deflection and oscillation of the tower. Further, in this way the choice of control strategy may also depend on the control history, or on for instance the actual operational parameter compared to the optimal operational parameter as determined by the main control system of the wind turbine.

According to another embodiment of the invention, the at least one control strategy of the set comprises keeping the actual operational parameter unchanged, which strategy may be advantageous if the actual operating operational parameter falls within the space of acceptable operating situations and if the operational parameter is already in accordance with its optimal value as for instance pre-defined, chosen or determined by a main control system of the wind turbine.

According to a further embodiment of the invention, the at least one control strategy of the set comprises a relaxation strategy of the actual operational parameter towards the optimally parameter as determined by the main control system of the wind turbine. This strategy may be advantageous in situations where an active control is no longer needed in the actual operating situation (for instance because a wind gust has passed over), and the operational parameter therefore can be relaxed or gradually changed back towards the optimal operational parameter as e.g. determined or chosen in the main control system of the wind turbine.

According to a further embodiment of the invention, the at least one control strategy of said set comprises an activating strategy of the actual operational parameter away from the optimal parameter as e.g. determined by a main control system of the wind turbine. This activation strategy governs the situations of the actual operating situation falling outside the space or confidence volume of acceptable situations and is reacted to for instance by pitching one or more blades out of the wind. As the actual operational situation is dependent on the nacelle acceleration, velocity and relative position, the control method will detect and react to an unacceptable operating situation earlier and more precisely than conventional control systems often only considering the speed of the wind acting on the rotor.

According to a further embodiment of the invention, the control function of the operational parameter is a function of the actual operational parameter which is advantageous for instance during relaxation of the operational parameter in the case of returning to normal con- trol operation after previously having had reacted to some extreme and non-acceptable operating situation.

According to a further embodiment of the invention, the control function of the operational parameter comprises reducing the operational parameter according to a first order relaxation of the present operational parameter. This is advantageous in providing an effective yet not to abrupt relaxation strategy in at least the cases described above. Further, an exponentially decreasing relaxation strategy is simple and effective to implement in comparison to e.g. a linear decrease.

The invention in a further embodiment relates to a method according to any of the previous, where the operational parameter is a function of the actual operating situation relative to said space of acceptable operating situations. Hereby is obtained that the operational parameter may be changed in dependence of how close the actual situation is to becoming unacceptable or acceptable. This is advantageous in providing a gradual control more insensitive to any possible uncertainties or even errors in determining the confidence volume and/or the actual operating situation.

Finally, in a further embodiment of the invention, the control function of the operational parameter comprises setting the rate of the operational parameter equal to a predetermined constant such as for instance 0 or 10 degrees/s.

Brief description of the drawings

In the following different embodiments of the invention will be described with reference to the drawings, wherein:

Fig. 1 shows the tower base moments for a wind turbine for a series of different wind and production situations,

Fig. 2 shows a simulation of a wind speed scenario comprising a gust, the resulting pitch modification caused by that wind and according to a conventional pitch regulation scheme, and the nacelle displacement resulting herefrom,

Fig. 3 illustrates the derivations of the nacelle velocity from the nacelle acceleration in a continuous-time and a discrete-time setting, respectively, Fig. 4 illustrates the deviational position of the nacelle from the nacelle acceleration in a continuous-time and a discrete-time setting, respectively,

Figs. 5 illustrates a number of normal operating situations in different views in the (acceleration, velocity, position)-space and the chosen acceptable operating spaces derived therefrom,

Fig. 6 illustrates the method for controlling the operational parameters of the wind turbine according to one embodiment of the invention,

Fig. 7 illustrates the reasoning for applying the additional condition of positive derivative of the r-parameter wrt the acceleration before an operation parameter such as the pitch angle is changed positively,

Fig. 8 and 9 illustrates the resulting regulation scheme according to embodiments of the invention for two different wind speed scenarios and as compared to a conventional control scheme,

Fig. 10 illustrates the pitch change as a function of the r-parameter for three different applicable embodiments according to the present invention,

Fig. 11 illustrates a method for controlling the operational parameters of the wind turbine during shutdown according to one embodiment of the invention,

Fig. 12 illustrates the control scheme during shutdown in the nacelle (velocity, accelerationn) space, and

Fig. 13 shows a simulation of a wind speed scenario comprising a gust, the resulting genera- tor speed, the pitch modification caused by that wind, and the nacelle displacement resulting herefrom according to a conventional pitch regulation scheme and according to a proposed regulation scheme, respectively.

Detailed description of the drawings

Figure 1 shows the maximum tower base bending moment M, 101 for a wind turbine plotted against the mean wind speed v, 102 for a series of different design wind and production situations. These design load cases cover normal production at normal wind speeds, extreme gusts of wind, grid outage, startup with wind gusts, shutdown with gust, and 50-year storm (stand-still). Looking at the specific load cases, a quite distinct grouping of events as outlined by the ellipsis 103, 104, 105 in the figure is observed. The load cases encircled by the dotted ellipse 103 represent loads obtained by simulating idling in a 50-year storm. Often, these loads are referred to as "stand-still" loads. As these loads do not occur during production, these loads cannot be reduced by applying extra features in the control system.

The load cases encircled by the dashed ellipse 104 represents extreme loads occurring during rare events such as wind gusts, extreme operating gusts occurring during emergency shutdowns, gusts occurring during pause-run or pause-run, grid errors, etc.. Notice that these loads exceed the 50 year stand-still loads on the right. Considerable load reductions can therefore be achieved if the rare events in the dashed ellipse 104 are mitigated by control features. Potentially, the maximum base moment could be reduced by app. 40%.

The remaining load cases encircled by the solid ellipse 105 represent all other load cases not driving the tower design with respect to extreme loads, including normal production.

Although the sizes of the extreme loads and their exact dependence of the wind speed of course vary with the specific product type of the wind turbine, the general pattern of extreme tower bending moments in the partial load operation mode at lower wind speeds and primarily caused by rare events in the wind is the same as in figure 1.

The extreme tower bending M occurs in the forward (up against the wind direction) as well as the backward direction (downwind). When the blades are pitched out, the aerodynamic damping in the fore-aft direction is poor. Therefore, if the tower is deflected forwards (due to e.g. an uncontrolled stop exerting negative thrust), a backward deflection of significant magnitude will follow soon after. In general the backward deflections (negative base moments) are caused by gusts whereas the forwards deflections (positive base moments) partly are caused by negative thrust due to pitching.

This is illustrated in figure 2 where is shown the reaction of a conventionally pitch-regulated wind turbine to a wind field comprising a wind gust 210. The first curve shows the wind speed V ₁ 102 as function of the time t, 204. The time t out of the x-axis is the same for all three curves, 201, 205, and 206. A gust 210 occurs around t=60 s, starting with a relative small drop in velocity as characteristic for a gust. In the simulation, the results of which are depicted in the figure, the wind turbine is an active pitch regulated turbine, and the blades are pitched θ, 203 in as seen from the middle curve 205 as a reaction to the initial decrease in wind speed in the wind gust, however a little delayed. As the wind speed V then increases significantly from the gust 210, the blades are pitched out of the wind, but this happens too late to avoid firstly the tower bending downwind (as can be seen from the curve 206 showing the tower downwind displacement x, 200). The wind gust lowering off results in the conventional pitch controller pitching in (at approx. t=64s) just when the tower begins it backwards motion. This results in excessive backward bending of the tower peaking at around t=68s as marked with the inserted circle 208. Summarizing, the incapability of the conventional pitch controller to react instantaneously to the wind speed and its limitation to take only the current wind speed into account renders the control system incapable of reacting to dynamic effects in the wind such as wind gusts or turbulence. These operation conditions although extreme and rare result in excessive loads on the tower, leading to heavier tower constructions with increased use of material in order to uptake the excessive loads and bending mo- ments.

The basic idea behind the control method according to embodiments of the present invention to be described in the following is to refine the controlling of the wind turbine by taking the actual operation situation of the wind turbine into account and not only the wind speed as in conventional control strategies.

The parameters describing the operation situation comprises the acceleration a of the nacelle, the nacelle velocity v, and its position. The nacelle acceleration a(t) is measured by an accel- erometer placed in or on the nacelle or near the top of the wind turbine tower and measuring the fore-aft acceleration of the nacelle. The accelerometer could also be placed approximately halfway up the tower height thereby being influenced minimally by the 2 ^nd order vibrations of the tower. The acceleration could also be measured by means of a two-way accelerometer from which the fore-aft acceleration could be derived based on knowledge on its orientation and position. Alternatively the size of the acceleration-vector (the mean square value) could be used in the derivations for the other parameters and in the expressions for the control system as described in the following.

The fore-aft velocity v(t) of the nacelle can be derived from the measured acceleration a(t) by integrating the measured acceleration signal. Pure integration will however provide a velocity estimate offset by the velocity at time t=0, which is unknown. This constant offset v(0) is removed by filtering the integrated acceleration signal with an ideal high-pass filter whereby only the DC component or mean value is removed. The filter will also remove the mean value of the true velocity v{t), which however is not a problem, since the mean nacelle velocity will approach zero as time goes by. Thus, the ideally high-pass filtered and integrated accelerometer signal provides a velocity estimate that will converge towards the true velocity.

The two block diagrams in figure 3 depict the principle of the derivation of the velocity 400 in a continuous-time setting and a discrete-time setting, respectively. In the continuous-time setting shown in figure 3A the acceleration signal a(t), 300 - possibly perturbed by an offset c, 301 - is fed to a DC offset removal filter, 303. The output from this filter is then fed to an integrator, 304. In one embodiment, the DC offset filter has an asymptotic slope of 40 dB/decade thereby ensuring that the resulting transfer function from input to output has a zero gain at 0 Hz. The two steps of the DC offset removal filter, 303 and the integrator, 304 corresponds to a band-pass filter, 305. The velocity 400 may similarly be estimated in a discrete-time setting as illustrated in figure 3B where a second order band-pass filter is employed which may be interpreted as a DC-offset removal filter followed by a leaky integrator. This could for example be a discrete time filter with a reference point in zero and a double pole close to I ₁ such as e.g. 0.95 or even 0.99. To place the double pole relatively close to 1 is advantageous by preventing any relevant low frequency components in the acceleration signal to be filtered off.

The procedure described above can be used for the nacelle position estimation as well. As acceleration is the 2 ^nd derivative of the position, an estimate of nacelle position can be ob- tained by integrating the acceleration a(t), 300 twice. Basically, the velocity estimate is integrated and DC-offset filtered to yield the position. As was also the case for the nacelle velocity estimation, the initial conditions for position and velocity are unknown, and one needs to address the problem with DC offsets in the acceleration measurement. In the case of velocity estimation the fact that the nacelle mean velocity is zero was exploited. This is however not the case for the position. As a consequence, only a relative or a deviational position can be obtained. Thus, the algorithm will provide the tower position x(t), 200 relative to a running mean, and not an absolute position. In other words, only the AC component of the position can be estimated.

In a continuous-time setting, an implementation is outlined in figure 4A comprising a first step of feeding the acceleration a(t), 300 to a DC offset removal filter, 303. The filter need not be the same as employed in the velocity estimate, figures 3A and B. Thereafter the signal is fed to a double integrator, 401. The two steps correspond to a band-pass filter, 402.

A discrete-time implementation is depicted in figure 4B, leaving the position estimator as a 3 ^rd order digital filter.

The measured nacelle acceleration a(t), 300, and the nacelle velocity v(t), 400, and relative position of the nacelle x(t), 200 being determined from the acceleration, are as mentioned above used to describe the actual operating situation of the wind turbine. In one embodiment of the invention, the nacelle acceleration is measured on a continuous basis. When these quantities or a certain combination hereof (expressed as the actual operating situation) is exceeded, the control feature according to the invention is activated and one or more of the blades are effectively pitched out of the wind or by other means controlled in order to avoid the extreme loads on the tower and dampening the tower vibrations.

The rare, extreme, and potentially dangerous combinations of acceleration, velocity, and deviational position are identified by first determining the space or spaces 500 of acceptable operating situations for the wind turbine based on sets of normal operating situations 501. In figure 5 scatter plots of deviational position x, 207, velocity v(t), 400, and acceleration a(t), 300 for normal production are showed in different views,- 3-dimesionally (upper plot to the left), and as seen along the axis of the acceleration (top right), velocity (bottom left), and relative position (bottom right), respectively. Clear correlations between the parameters in these normal operating situations are evident from the plots. For instance a correlation between the deviational or relative position 200 and the nacelle acceleration 300 is observed on the lower left plot where the slope 502 of the ellipse 500 is governed by the stiffness of the tower. Thus, a linear fit would reveal a first-order term approximately equal to the squared eigenfrequency. The normal operating situations 501 plotted in figure 5 are based on simu- lated data from numerous series of different test wind conditions and production situations. The normal operating situations 501 could also be based on or supplemented with data collected from the same or other wind turbine operating under normal wind conditions. Similarly, the points of normal operating situations 501 could be based on statistical data.

Simulations show that standard deviation of the nacelle acceleration, velocity, and relative position increases with wind speed. Therefore, it is reasonable to base any threshold values on the worst-case quantities of the wind speed thereby providing a very cconservative control strategy. In this way it is ensured that the production is not affected during production in normal wind turbulence but only the rare situations of e.g. extreme wind gusts. This strategy leaves normal production practically unaffected. That is, during production at normal turbu- lence levels, the control strategy according to the present invention should remain inactive to preserve energy production. The scatter plots depicted in figure 5 are all determined from simulations at a mean wind velocity of V=24 m/s which for many wind turbine is the maximally allowed wind speed prior to shut-down or stop of the wind turbine, therefore resulting in a very conservative description of the space for acceptable operating situation.

From these sets of normal operating situations are determined the function or space of acceptable operating situations 500 defining the normal combinations of acceleration, velocity, and relative position as those point lying within the space of confidence volume 500, in contrary to those points lying outside which are deemed abnormal and unacceptable. In this sense the functional expression for the space of acceptable operating situations corresponds to a threshold function at least partly determining the following control strategy. In the pre- sent embodiment shown in figure 5 the space of acceptable operating situations 500 is chosen to be given by a function f(x, v, a) =1, describing an ellipsoid.

In other embodiments of the invention, the space 500 of acceptable operating situations could be determined by surface fitting to the data set of normal operating situations. The confidence volume 500 could also be expressed by a number of functions, or by any parametric functions.

In other embodiments of the invention, the normal operating situations could be collected, determined, or even refined for different mean wind speeds or for different predefined ranges of mean wind speed. This could result in different spaces of acceptable situations with their sizes, shapes, and orientations varying with the actual wind velocity.

The steps of the method for controlling the operational parameters of the wind turbine according to one embodiment of the invention are illustrated in figure 6. The nacelle acceleration a(t), 200 at a given time t is measured as described previously, and the nacelle velocity v(a(t)), 400 and the deviational position of the nacelle x(a(t),v(t)), 200 are determined from the acceleration. The actual operational situation governed by these parameters is then in the step 601 inserted in the function r ² =f(x(t), v(t), a(t)) where the function f describes the space 500 of acceptable operating situations which is pre-determined from a set of normal operating situations 501. The statistical r-parameter then expresses the distance from the actual operating situation (the (x(t), v(t), arø>point) to the confidence volume governed by f=l . In a simple embodiment of the control method, the control of the operational parameter is activated when r>l, corresponding to the actual operating situation falling outside the space 500 of acceptable operating situations. This leads to an activation strategy S_l, 603 being chosen which in one embodiment comprises setting the rate of the operational parameter equal to some constant. In the case of pitch regulating, the pitch rate may in one em- bodiment of the invention be chosen equal to a constant between 1 and 20 degrees/s, preferably between 5 and 15 degrees/s, and in a specific embodiment equal to 10 degrees/s. In another embodiment of the invention the activation strategy S_l comprises setting the rate of the operational parameter (for instance the pitch rate) proportional to the nacelle velocity.

The threshold function can be combined with other criteria before the control strategy is cho- sen. This is performed in the method step 604 in the figure 6. In one embodiment of the invention these additional criteria which must be fulfilled before the activating of the operational parameter is chosen, comprises only changing the operational parameter in case the nacelle velocity v(t), 400 is positive or the derivative of the parameter r with respect to the r)r ² nacelle acceleration a(t) is positive, > o ■ The reasoning for these specific criteria is fur- Da ther elaborated below in relation to figure 7.

Other additional activation conditions for the control method may be evaluated such as specifying a maximum or minimum operational parameter for which the control method is deacti- vated. This may be advantageous in preventing the turbine from stopping completely due to the control method being repeatedly activated by small-amplitude oscillatory motions observed in the later stages of control sequences otherwise potentially leading to complete stops.

If the control system is disabled because the actual operating situation again falls within the space 500 of acceptable operating situations, i.e. r<l, then the control strategy S_2, 606 may be chosen in which the operational parameter may be decreased through e.g. a first order asymmetric filter or in another relaxed or gradually changed back towards the optimal operational parameter θ _op t being optimal for the current wind speed and production condition and as e.g. determined by the main control system of the wind turbine.

Finally, the operational parameter is kept unchanged corresponding to the third control strategy S_3, 607 if the actual operating situation is acceptable and the actual operational parameter has not been changed in the previous step by the control method, i.e. the actual operational parameter is equal to the optimal parameter θ _opt , as evaluated in the step 608.

In the case of blade pitch regulation, the control method described above can be summarized as choosing a pitch rate proportional to the nacelle velocity or equal to a constant once an excessive excitation has been detected, thereby effectively pitching until the nacelle deflection has stopped. For negative nacelle velocities (forwards motion), the pitch is held constant to cater for a conservative scheme. When the control method is disabled, the pitch angle is decreased through a first order asymmetric filter.

Figure 7 illustrates the reasoning for applying the additional condition of positive derivative of the operational parameter r with respect to the acceleration, before an operation parameter such as the pitch angle is changed positively. First, it is apparent that the thrust force on the rotor, and, in turn, the acceleration of the nacelle will decrease with increasing pitch angle. r)r ²

Thus, we can conclude that if the derivative of r with respect to a is positive, > o _/ we

9a activate the operational parameter control and pitch out, effectively reducing the thrust force and acceleration, and, in turn, bringing the operating trajectory closer to space or confidence volume 500 representing normal operation. This is the case for the operating point 701 marked in the figure. However, if the derivative is negative, pitching out would move the operating situation illustrated by the point 702 in the figure further away from the normal volume 500. This could be the case in situations where the extremely low thrust is experi- enced as a result of steeply falling wind. In such situations, pitching out would just make things worse.

Then, considering the nacelle velocity, if the velocity is positive, a decrease in the thrust force will decrease the positive work done on the system, thus decreasing the oscillation level. Therefore, the control method is activated for positive nacelle velocities, v>0.

Figure 8 and 9 illustrates the effect of the control method applied to pitch regulation for two different simulated wind speed scenarios and compared to a conventional pitch control. The wind speed scenario in figure 8 is the same as shown in figure 2, where the wind field comprises a gust 210. The top curve 210 shows the wind speed V ₁ 102 as function of the time t, 204. The blade pitch angle θ, 203 and the tower deflection x, 200 is shown in the middle and lowermost, respectively, and as resulting from a conventional pitch control scheme 205, 206 in dotted lines, and as a result of having applied a control method according to the present invention, 801, 802 in solid lines. As can be seen the two pitch regulation schemes are identical at first, until the normal controller 205, 206 pitches into the wind to obtain nominal generator speed and not taking the tower deflections into account. As also discussed in figure 2, this leads to excessive backwards bending of the tower at the approximate time of t=68s.

The control system according to the invention detects the gust and starts pitching out around t=62s. The control system further reacts a second time around t=66s. In the time intervals 800, the control method is disabled again and the pitch angle is decreased through a first order function towards the pitch angle that would have been optimal in the given wind situa- tion had it not been for the presence of the wind gust 210. As can be seen by comparison of the two curves for the deflection, 206 and 802, the proposed control method leads to a far more effective damping of the oscillations of the tower, lower extreme tower deflections, and hence reduced tower bending moments.

In the simulation shown in figure 9, several gusts occur and the wind field comprises more turbulence as can be seen from the top curve 901 of the wind speed V ₁ 102 as a function of the time t, 204. After the first gust, the normal controller (dotted curves, 902, 903) pitches into the wind to obtain nominal generator speed and does not take into account the heavy tower oscillation. Therefore excessive backwards bending is observed at approximately t=30s. The control method according to the invention on the other hand (solid lines, 904, 905) detects the tower oscillation and applies the defensive strategy of increasing the pitch angle until the nacelle velocity is zero. In another embodiment of the invention, the operational parameter may be changed not only depending on whether the operating point is inside or outside the space 500 of acceptable operating situations but also as a function of how close the actual operating situation is to the surface or space 500. As described previously, the statistical r-parameter (r ² =f(x(t), v(t), a(t))) expresses the distance from the actual operating situation (the (x(t), v(t), a(t))-po\r\t) to the confidence volume governed by f=l. The rate of the operational parameter, 1000

(such as the pitch rate (P) may therefore advantageously be changes as a function of the r- parameter. This is illustrated in figure 10 for three different functions, 1001, 1002, 1003. Here the step-function 1001 corresponds to the previously described embodiment of not changing the operational parameter ((P=O), if the operating situation lies within the confidence volume 500 (r< l). The function 1002 illustrates a pitch change which is linearly dependent on r, i.e. the distance of the actual operating point to the confidence volume 500. The pitch change is increased the further away the point is away from the acceptable space, but the functional dependence also up for changing the operational parameter already as the operating situation approaches the surface (r->l). This is also the case for the function 1003 in a third embodiment. Here, the change of the operational parameter rate is also changed as a function of the r -parameter. In this latter case, the operational parameter may even be changed aggressively towards the space of acceptable operating situations for very low values of the r-parameter (negative rate) expressing that the actual operating situation is rela- tively well within or even close to the center of the confidence space 500 corresponding to little or no oscillation of the tower and therefore a wind field of no or low turbulence. The control of the operational parameters may in this way be tuned towards increasing the power production in stable wind conditions or wind fields of low turbulence.

Other functions than those depicted in figure 10 may of course be chosen to obtain the de- sired active control of the operational parameters.

Generally, another advantage by the aforementioned control methods is that the number of overspeed situations will be reduced considerable because the control method reacts quickly to gusts.

Some or all of the above described features may also be used in a control scheme during stop sequences on the rotor where the blades are pitched out of the wind towards feather. Normally this procedure inherently reduces the thrust force on the rotor dramatically in a short time, and, in most cases even exposes the rotor to negative thrust. That is, the wind will actually pull the turbine as opposed to normal operation, where the wind exerts a force on the rotor, effectively deflecting the tower backwards. The thrust decrease will reduce the tower deflection, causing the nacelle to have negative velocity (moving forwards). When negative thrust is applied while the nacelle is moving forwards, the thrust will do positive work on the structural system. As the tower/nacelle is a poorly damped system, the combined effect on the forwards motion and negative thrust is, in some cases, extreme deflection of the tower in the forwards direction. In some cases, the resulting loads are design-driving. Therefore, mitigation of the loads resulting from such stops could decrease the characteristic load governing the design loads, effectively reducing the cost price of the turbine towers. Thus, often extreme loads occur when feathering causes excessive oscillatory motion of the turbine structure.

Such extreme loads can be reduced if not obviated by the application of some or all of the previously described steps used in the general control method of reducing extreme tower loads.

In one embodiment, illustrated in figure 11, similarly to the aforementioned control scheme described in relation to figure 6, the nacelle acceleration a(t), 200 at a given time t is measured, and the nacelle velocity v(a(t)), 400 is determined from the acceleration. Further, the deviational position of the nacelle x(a(t),v(t)), 200 may optionally also be determined and used in the evaluation of the actual operational situation and could in some situation increase the accuracy. However, the nacelle position is not a requisite parameter in the stop control. The nacelle acceleration and velocity is then in the step 1101 inserted in the function ! ² =g(v(t), a(t)) where the function g describes the space 500 of acceptable operating situa- tions which is pre-determined from a set of normal operating situations 501 in the (velocity, acceleration)-space. In a simple embodiment of the stop control method, the feathering of the blades (e.g. the pitch continuously increased by a constant pitch rate) is maintained when r<l, and corresponding to the actual operating situation falling within the space 500 of acceptable operating situations, 1102. On the other hand, if the actual operating situation falls outside the space 500 of acceptable operating situations (r>l) and the nacelle velocity is negative, v(t)<0, 1103, then the stop operation is deactivated or paused, 1104, and the blade pitch kept constant until the velocity becomes positive or the actual operating situation again falls within the space 500 of acceptable operating situations. In other words the stop operating scheme can be loosely formulated as monitor the tower oscillation level during shutdown and if this oscillation level exceeds a predefined threshold then avoid making the situation worse by disallowing pitching. The scheme is illustrated graphically in figure 12.

A simulation shown in figure 13 illustrates the above described control scheme during shutdown or at least a partial stop of the turbine. The curves show from above and down, the wind speed V ₁ 102, the generator speed G, 1300, the blade pitch angle θ, 203 and the tower deflection x, 200 all as a function of the time t, 204. Further, is shown the mentioned parameters as resulting from a conventional pitch control scheme 1301, 1302, 1303 in dotted lines, and as a result of having applied a stop control method according to the present invention, 1304, 1305, 1306 in solid lines. As can be seen the two pitch regulation schemes are identical at first, until the control method according to the invention causes the pitching 1305 to be disrupted and paused for a period of time 1310 whereby the excessive bending of the tower is avoided instead yielding a far more effective damping of the oscillations of the tower, lower extreme tower deflections 1306, and hence reduced tower bending moments.

The above proposed control method has the advantages of not requiring a full pitch control as in normal operation. The control system actuators may be restricted to be on/off valves in the pitch hydraulics able to interrupt the normal feathering sequence. Further, the use of sensors can be limited to an accelerometer working in the fore-aft direction. Also, measurements of the rotational speed are not needed, which is advantageous in that reliable estimates of the overall rotational speed of the rotor are difficult to obtain for one reason because the rotational speed observed in the hub is heavily perturbed by edgewise/drive-train vibrations.

While preferred embodiments of the invention have been described, it should be understood that the invention is not so limited and modifications may be made without departing from the invention. The scope of the invention is defined by the appended claims, and all devices that come within the meaning of the claims, either literally or by equivalence, are intended to be embraced therein.