Method and control system for reducing the fatigue loads in the components of a wind turbine subjected to asymmetrical loading of the rotor plane

[0001] The present application is a continuation of pending International patent application PCT/DK2006/000153, filed Mar. 16, 2006, which designates the United States, the content of which is incorporated herein by reference.

[0002] The present invention relates to a method for reducing fatigue loads in the components of a wind turbine, a control system for reducing the fatigue loads in the components of the wind turbine subjected to asymmetrical loading of the rotor plane, a wind turbine and a wind park.

[0003] Wind turbine controllers have been deployed in wind turbines for years with the purpose of controlling the overall power output.

[0004] The power output from a modern wind turbine can be controlled by means of a control system for regulating the pitch angle of the rotor blades. The rotor rotation speed and power output of the wind turbine can hereby be initially controlled e.g. before a transfer to a utility grid through power converting means. An advantage of this control is a protection of the rotor from rotating at an excessive speed at high wind speeds and save the rotor blades from excessive loads.

[0005] Especially for large rotor diameters, the distribution of the wind inflow profile can be non-uniform over the area of the rotor, resulting in a non-uniform load to each rotor blade as a function of one full rotation, as well as asymmetrical out of plane loadings for the drive train of the wind turbine. For a free wind inflow situation the wind shear distribution is approximately linear and the said load as a function of rotation is of nearly sinusoidal behavior with a frequency equal to the rotor rotation frequency. In order to keep a more constant load on the rotor blades, pitch control functions have been applied to wind turbine pitch controllers, where a rotor-cyclic correction with a frequency equal to the rotor rotation has been added to the overall pitch angle setting of the individual rotor blades.

[0006] Any obstacles within certain up wind distance of a wind turbine create a wake for the wind turbine and consequently eliminate the free wind inflow situation. An example of an obstacle may be other wind turbines, as a wind turbine always cast a wake in the downwind direction.

[0007] Especially in wind parks this fact may significantly influence the inflow on wind turbines situated in the down wind direction. This results in a more complex wind shear distribution than compared to a free wind inflow situation. The said complex wind distribution profile can result in wind turbulence and in turn fluctuating fatigue loads on the wind turbine components. So in order to avoid too much wind turbulence around the turbines, downstream wind turbines are spaced relative far apart resulting in very area consuming wind parks.

[0008] It is therefore an object of the present invention to provide a method and a technique that allows an improved wind turbine control strategy in relation to more complex wind shear distribution.

[0009] The invention provides a method for reducing fatigue loads in the components of a wind turbine subjected to asymmetrical loading of its rotor, comprising the steps of:

• • [0010] repeatedly collecting and storing load data of the rotor,
  • [0011] determining a load distribution function for the rotor from said stored data,
  • [0012] deriving a plurality of periodic functions from said load distribution function,
  • [0013] determining actions for the wind turbine control means for reducing the fatigue load of the wind turbine components from said derived plurality of periodic functions, and
  • [0014] implementing of said determined actions on the wind turbine control means.

[0015] Hereby it is possible to reduce the fluctuating loads on the wind turbine components which facilitate less mechanical wear and tear on the most load-exposed parts of the wind turbine which in turn results in less required service, fewer break-downs and a prolonged life-span for the wind turbine. Alternatively it is hereby possible to increase the power capture of the wind turbine and maintain a low load on the rotor or combinations hereof.

[0016] In an aspect of the invention, said derived plurality of periodic functions is sinusoidal and/or cosinusoidal functions. By using derived sinusoidal and/or cosinusoidal functions, it is possible to approximate a pitch correction in close proximity of the desired.

[0017] In another aspect of the invention, the frequencies of said plurality of periodic functions is limited series of different integer multiples of the rotor frequency e.g. up to four times the rotor frequency such as any of first, second, third and fourth multiply of the rotor frequency or combinations of at least two of said multiplies. By using limited series of periodic functions it is possible to quickly establish a pitch correction in close proximity of the desired, without requiring significant computational power.

[0018] In another aspect of the invention, at least an amplitude component and a phase component are determined for each of said derived periodic functions. Determining said amplitude and phase components facilitates the overall data processing in the computing means and facilitates the link in time and azimuth location between the asymmetrical rotor load and the corrective pitch action.

[0019] In another aspect of the invention, said plurality of periodic functions includes at least one function with a frequency equal to the rotor frequency. By including periodic functions with a frequency equal to rotor frequency, it is possible to approximate a pitch correction close to the desired, simplified in a free or near free wind inflow situation.

[0020] In another aspect of the invention, said plurality of periodic functions includes at least one function with a frequency equal to four times the rotor frequency. By including periodic functions with a frequency equal to four times the rotor frequency, it is possible to approximate a pitch correction close to the desired, simplified in a partly or full wake situation.

[0021] In another aspect of the invention, said plurality of periodic functions are derived by means of a Discrete Fourier Transform applied to the load distribution function. By using Discrete Fourier Transformation to derive said plurality of periodic functions, it possible to use well known, fast and liable programming techniques to implement the program code on the computing means.

[0022] In another aspect of the invention, said plurality of periodic functions are derived by means of inverse relations between first periodic harmonic and measured blade modal amplitudes.

[0023] In another aspect of the invention, said plurality of periodic functions are derived by means of successive band-pass filtering applied to the load distribution function.

[0024] In another aspect of the invention, said plurality of periodic functions are derived by means of a Recursive Least Square estimator applied to the load distribution function.

[0025] In another aspect of the invention, said load data are collected by measuring the blade root bending moments whereby a representative measurement can be obtained preferable with already existing sensor means.

[0026] In another aspect of the invention, said bending moments are measured for at least one blade. By doing measurements on one blade only, data can be used to establish a optimized pitch correction applied to all blades, assuming that all rotor blades undergoes the same asymmetrical load as a function of one full rotor turn.

[0027] In another aspect of the invention, said bending moments art measured for more than one blade e.g. two blades of the wind turbine. By doing measurements on more than one blade, optimal individual pitch correction can be applied to each individual blade.

[0028] In another aspect of the invention, said bending moments are measured in two substantially perpendicular directions.

[0029] In another aspect of the invention, said load data are collected by measuring the angle of attack for the blades whereby a representative measurement can be obtained preferable with existing measuring means.

[0030] In another aspect of the invention, said load data are collected by measuring the forces on a wind turbine main shaft such as a low or high speed shaft.

[0031] In another aspect of the invention, said load forces on said shaft are measured in two substantially perpendicular directions.

[0032] In another aspect of the invention, said wind turbine control means comprises a blade pitch control mechanism in order to be able to implement said determined actions.

[0033] In another aspect of the invention, said load forces are measured continuous or for a predetermined period of time, depending on the degree and speed of variations in the wind inflow situation and on the necessity for measurement and control.

[0034] In another aspect of the invention, said predetermined period of time equals 0.01 to 0.5 full rotations of the rotor, preferably in the range from 0.1 to 0.3 full rotations of the rotor depending on necessity here for.

[0035] In another aspect of the invention, said predetermined period of time equals 0.5 to 6 full rotations of the rotor, preferably in the range from 0.75 to 3 full rotations of the rotor depending on necessity here for.

[0036] The invention also relates to a control system as well as a wind turbine and wind park.

[0051] FIG. 1 illustrates a modern wind turbine 1 with a tower 2 and a wind turbine nacelle 3 positioned on top of the tower.

[0052] The wind turbine rotor, comprising at least one blade such as three wind turbine blades 5 as illustrated, is connected to the hub 4 through pitch mechanisms 16. Each pitch mechanism includes a blade bearing and pitch actuating means which allows the blade to pitch in relation to the wind. The pitch process is controlled by a pitch controller as will be further explained below.

[0053] The blades 5 of the wind turbine rotor are connected to the nacelle through the low speed shaft 4 which extends out of the nacelle front.

[0054] As illustrated in the figure, wind over a certain level will activate the rotor and allow it to rotate in a perpendicular direction to the wind. The rotation movement is converted to electric power which usually is supplied to the transmission grid as will be known by skilled persons within the area.

[0055] FIG. 2 illustrates how the Azimuth angle Ψ is measured as the angle between a virtual vertical line thru the centre of the low speed shaft 4 and a virtual line defined by the two endpoints: a—the centre of the low speed shaft 4a, and b—the tip point of the rotor blade 7. The Azimuth angle is measured for one reference rotor blade e.g. blade 1 as a function of time and position.

[0056] FIG. 3a illustrates one rotor blade 5 of a wind turbine connected to the nacelle 3 trough the low speed shaft 4 which extends out of the nacelle front.

[0057] The rotor blade is loaded by a wind force F_load(t) dependent of e.g. the wind direction relative to the rotor blade, the area of the rotor blade, the pitch of the rotor blade etc. The said wind force which literally tries to break off the nacelle from the tower or the foundation produces a load bending moment m_x in the low speed shaft 4 and in the root of rotor blade 10 around its centerline 8.

[0058] FIG. 3b illustrates a formalized diagram of the in situ forces acting on one rotor blade illustrates the center point of the low speed shaft 4a, the horizontal centerline of the low speed shaft 8a, the vertical centerline of the rotor blade through the center point of the low speed shaft 9, a summarized wind force F_load(t) and the direction of the load bending moment (or out of plane moment) m_x of blade number x.

[0059] FIG. 4 illustrates schematically a preferred embodiment of a control system for controlling the pitch angles of the wind turbine blades

[0060] Data of the wind turbine 1 are measured with sensor means 11 such as pitch position sensors, blade load sensors, azimuth sensors etc. The measured sensor data are supplied to computing means 12 in order to convert the data to a feedback signal. The feedback signal is used in the pitch control system 13 for controlling the pitch angle by establishing control values for controlling said at least one wind turbine blade 5.

[0061] The computing means 12 preferably includes a microprocessor and computer storage means for continuous control of the said feedback signal.

[0062] By continuously measuring the present load moments values on the rotor-blades, calculating an desired optimal pitch angle setting and feeding this information to the pitch control system in a closed feedback loop it is possible to optimize the control values to (substantially) to control the rotor at the design limits of the wind turbine and especially the design limits of the wind turbine blades.

[0063] An example of prior art for controlling out of plane moment loads on wind turbine blades of a wind turbine is here described.

[0064] The blade root loads M_R=[m₁ m₂ m₃]^T on the rotor blades of a 3 bladed wind turbine are defined as a result of a given linear wind shear distribution between a rotor blade top position (ψ=0) and down-ward position (ψ=π) corresponding closely to an idealized free wind inflow situation.

[0065] FIG. 5a illustrates a typical picture of said moments for free inflow conditions.

[0066] Transforming M_R into a coordinate system defined by the tilt, yaw and thrust equivalent directions, the respective moments loads m_tilt, m_yaw, m_sum become:

[0067] $m_{\mathrm{tilt}}=m_1·\cos \left(\Psi \right)+m_2·\cos \left(\Psi +\frac{4}{3}\pi \right)+m_3·\cos \left(\Psi +\frac{2}{3}\pi \right)$ $m_{\mathrm{yaw}}=-m_1·\sin \left(\Psi \right)-m_2·\sin \left(\Psi +\frac{4}{3}\pi \right)-m_3·\sin \left(\Psi +\frac{2}{3}\pi \right)$ $m_{\mathrm{sum}}=m_1+m_2+m_3$

[0068] For the loads illustrated in FIG. 5a, the said transformed moment loads, m_tilt, m_yaw, are illustrated in FIG. 5b as a function of one full rotation of the rotor. For this idealized example of a prior art, m_tilt, m_yaw are constant.

[0069] The close to sinusoidal behavior of M_R as illustrated in FIG. 5a will result in fatigue loads on the rotor blades. A technique to partly compensate for these altering loads on the rotor blades can therefore be to individually control the rotor blades during a full rotation of a blade in order to level the distribution of wind forces i.e. a rotor blade is pitched less into the wind at the top than at the bottom of the rotating movement performed by the rotor including the blades.

[0070] Due to this close relation between M_R and the desired controlling of the pitch angle, the desired pitch control signal is also a function of the Azimuth angle i.e. a sinusoidal function on a frequency equal to the rotor-rotation frequency. This technique is called cyclic or rotor-cyclic pitch of the wind turbine blades i.e. a cyclic change of the pitch angle during a full rotation of a blade.

[0071] When the rotor blade enters a wake it is exposed to a step-like shearing force. This has been confirmed by actual measurements on wind turbines. Still in order to keep a constant load on the rotor blades under this condition, said rotor-cyclic pitch control can be applied resulting in a basic optimization of the load. But as the affected load is of step-like behavior and the said rotor-cyclic pitch control is of sinusoidal behavior there will always occur a non negligible alternating force on the rotor blades.

[0072] This is illustrated in FIG. 6 for said idealized half-wake situation. The curve 14 illustrates a desired abrupt change in pitch angle control and the curve 15 illustrates an actual corrective pitch angle control applied by the said rotor-cyclic pitch technique. Due to the difference between the two curves, an angle error 16 is introduced still resulting in a possibility of increased fatigue loads on the rotor blades.

[0073] An example of the present invention for controlling out of plane moment loads on wind turbine blades of a wind turbine is here described.

[0074] FIG. 7 illustrates for the present invention a preferred embodiment of the said control system for controlling the pitch angles of the wind turbine blades.

[0075] The moment loads M_R=[m₁ m₂ m₃]^T on the rotor blades and the azimuth angle ψ is measured by the sensor means and feed to the computer means.

[0076] M_R is transformed into a coordinate system defined by the tilt, yaw and thrust equivalent direction M_F=[m_tilt m_yaw m_sum]^T=T·M_R

[0077] Where:

[0078] $T=\left[\begin{array}{ccc}\cos \left(\Psi \right)& \cos \left(\Psi +\frac{4}{3}\pi \right)& \cos \left(\Psi +\frac{2}{3}\pi \right)\\ -\sin \left(\Psi \right)& -\sin \left(\Psi +\frac{4}{3}\pi \right)& -\sin \left(\Psi +\frac{2}{3}\pi \right)\\ 1& 1& 1\end{array}\right]$

[0079] The inverse relation is given by: M_R=T⁻¹·M_F

[0080] M_F is data processed by a filter (H) to M_F^{h}, deriving and processing a plurality of harmonic functions on different multiple integers of the rotor frequency (ω_nom) in order to adapt the pitch angle control system to minimize the fluctuations on measured load data in such a way, that the loads on the rotor blades are kept constant or nearly constant.

[0081] A preferred embodiment of said data processing filter (H) is a Recursive Least Square (RLS) Estimator with exponential forgetting. This is a mathematical optimization technique that attempts to find a best fit to a set of data by attempting to minimize the sum of the squares of deviation between a set of observed data and a set of expected data.

[0082] The RLS processing algorithm is based on a few key-operators and can in a computer simulation be implemented after the following algorithm:

[0083]

φ ⁡ ( t ) = [ 1 ⁢ ⁢ cos ⁡ ( ω nom ⁢ t ) ⁢ ⁢ sin ⁡ ( ω nom ⁢ t ) ⁢ ⁢ cos ⁡ ( 2 ⁢ ω nom ⁢ t ) ⁢ ⁢ sin ⁡ ( 2 ⁢ ω nom ⁢ t ) cos ⁢ ( 3 ⁢ ω nom ⁢ t ) ⁢ ⁢ sin ⁡ ( 3 ⁢ ω nom ⁢ t ) ⁢ ⁢ cos ⁡ ( 4 ⁢ ω nom ⁢ t ) ⁢ ⁢ sin ⁡ ( 4 ⁢ ω nom ⁢ t ) ]
θ = [a0 a1 b1 a2 b2 a3 b3 a4 b4]T
R = 9 × 9 matrix initialized with zero elements
G = 9 × 1 vector initialized with zero elements
μ = 1/ko
for p = 1 . . . N (p is expressed as simulation step number, 1, 2, 3 . . . )
	t = p · Ts
	for i = 1 . . . 3 (iteration over mtilt myaw and msum)
	G(i) = (1 − μ) G(i) + μ φ(t) MF(i)
	R(i) = (1 − μ) R(i) + μ φ(t) φ(t)T
	θ(i) = (R(i))−1 G(i)
	MF(h)(i) = φ(t)Tθ(i)
End
end

[0084] In the above computer simulation example:

• • [0085] ω_nom=the nominal cyclic rotor frequency
  • [0086] φ=the harmonic analysis vector (here including components up to the 4^th harmonic)
  • [0087] θ=the harmonic amplitudes
  • [0088] R=is a 9×9 matrix, initialized with zero elements
  • [0089] G=is a 9×1 vector, initialized with zero elements
  • [0090] T_s=the simulation step time
  • [0091] μ=a forgetting factor
  • [0092] k₀=a positive integer defining the forgetting factor

[0093] It is worth noting that the said RLS filter is adaptive which yields that the output of the filter changes as a response to a change on the input.

[0094] A practical applied version of the data processing comprises computing means for digital data acquisition, harmonic analysis, RLS filter computation, data storage and D/A converting, continuously or for a predetermined period of time.

[0095] Due to time-delays in the sensor means, in the computer means and in the pitch control system, the corrective pitch angle control signal is time shifted in relation to the measured blade loads M_R. To correct for this, M_F^{h} is time shifted equivalently to synchronize i.e. M_FS^{h}=timeshift(M_F^{h})

[0096] A general time shift of a sum of harmonic signals can be realized as follows:

[0097] $s\left(t\right)=\sum _{i=1}^n{a}_i\cos \left({\omega }_{i}t\right)+b_i\sin \left({\omega }_{i}t\right)$ $\mathrm{so}$ $s\left(t+\tau \right)=\sum _{i=1}^n{P}^{\left\{i\right\}T}{C}^{\left\{i\right\}}\left(t\right)Q^{\left\{i\right\}}\left(\tau \right),\left(1\times 2\right)\times \left(2\times 2\right)=\left(1\times 1\right)$ $\mathrm{where}\text{:}$ $P^{\left\{i\right\}}={\left[a_i{b}_i\right]}^T$ $C^{\left\{i\right\}}\left(t\right)=\left[\begin{array}{cc}\cos \left({\omega }_{i}t\right)& -\sin \left({\omega }_{i}t\right)\\ \sin \left({\omega }_{i}t\right)& \cos \left({\omega }_{i}t\right)\end{array}\right]$ $Q^{\left\{i\right\}}\left(\tau \right)={\left[\cos \left({\omega }_i\tau \right)\sin \left({\omega }_i\tau \right)\right]}^T$

[0098] The filtered and time shifted signal M_FS^{h} is transformed from the fixed reference system back to the rotating reference system by M_R^{h}=T⁻¹·M_FS^{h}.

[0099] The signal M_R^{h} is multiplied with a gain for the conversion to radians i.e. β_dem^{h}=GainM_R^{h} and is added to the collective pitch demand signal β_dem^{c}.

[0100] FIG. 8a illustrates as an example moment loads M_R=[m₁ m₂ m₃]^T on the rotor blades of a 3 bladed wind turbine as a result of a horizontal step shear corresponding to an idealized half wake inflow situation.

[0101] Transforming M_R into a coordinate system defined by the tilt, yaw and thrust equivalent direction, the respective moments loads m_tilt, m_yaw, m_sum become:

[0102] $m_{\mathrm{tilt}}=m_1·\cos \left(\Psi \right)+m_2·\cos \left(\Psi +\frac{4}{3}\pi \right)+m_3·\cos \left(\Psi +\frac{2}{3}\pi \right)$ $m_{\mathrm{yaw}}=-m_1·\sin \left(\Psi \right)-m_2·\sin \left(\Psi +\frac{4}{3}\pi \right)-m_3·\sin \left(\Psi +\frac{2}{3}\pi \right)$ $m_{\mathrm{sum}}=m_1+m_2+m_3$

[0103] m_tilt, m_yaw, are illustrated in FIG. 8b as a function of one full rotation of the rotor.

[0104] Periodic functions like the functions illustrated in FIG. 8b can be resolved as an infinite sum of sines and cosines called a Fourier series and can in this case generally be expressed as:

[0105] $m\left(\Psi \right)=\frac{a_0}{2}+a_1\cos \left(\Psi \right)+b_1\sin \left(\Psi \right)+a_2\cos \left(2\Psi \right)+b_2\sin \left(2\Psi \right)+a_3\cos \left(3\Psi \right)+b_3\sin \left(3\Psi \right)\dots$ $\mathrm{where}\text{:}$ $a_i=\frac{1}{\pi }{\int }_0^{2\pi }m\left(\Psi \right)\cos \left(i\Psi \right)d\Psi$ $b_i=\frac{1}{\pi }{\int }_0^{2\pi }m\left(\Psi \right)\sin \left(i\Psi \right)d\Psi \mathrm{for}i=0,1,2,3,\dots$

[0106] The computation of Fourier series is known as Harmonic Analysis.

[0107] It is seen from the equation of m(Ψ) that the Fourier series consists of a non-alternating component, components that alternate according to the basic parameter Ψ and a plurality of periodic functions of different integer multiples of the basic frequency. The weighted Fourier coefficients a_i, b_i determine the amplitude of each harmonic frequency in the original signal.

[0108] The said RLS estimator data processes a truncated number of periodic functions derived by the harmonic analysis e.g. the first four multiple harmonics of the basic rotor frequency. The purpose of the RLS estimator is to produce an output signal that is feed to the pitch control system in order to minimize the energy in the load signal M_R i.e. to minimize the fluctuating loads on the rotor blades.

[0109] For this idealized example the input signals 17, 19 representing the loads moments m_tilt, and m_yaw of M_F respectively are illustrated in FIG. 9. The output signals m_tilt^{h} and m_yaw^{h} of M_F^{h} are represented by 18, 20 respectively. The said RLS filter has processed the first four multiple harmonics of the basic frequency.

[0110] The filtered signal M_F^{h} is time shifted to a signal M_FS^{h} and transformed from the fixed reference system back to the rotating reference system by M_R^{h}=T−1·M_FS^{h}

[0111] where:

[0112] $T=\left[\begin{array}{ccc}\cos \left(\Psi \right)& \cos \left(\Psi +\frac{4}{3}\pi \right)& \cos \left(\Psi +\frac{2}{3}\pi \right)\\ -\sin \left(\Psi \right)& -\sin \left(\Psi +\frac{4}{3}\pi \right)& -\sin \left(\Psi +\frac{2}{3}\pi \right)\\ 1& 1& 1\end{array}\right]$

[0113] Finally the filtered signal M_R^{h} is gain adjusted (β_dem^{h}) and added to an overall pitch angle control signal β_dem^{c} defined by a wind turbine speed controller and the summarized control signal β_dem is feed to the pitch controller that effectuates the desired actions.

[0114] FIG. 10 illustrates a data processed pitch control signal, e.g. m₁^{h} of M_R^{h}, from the example above, corresponding to an idealized half-wake situation. The curve 14 illustrates a desired abrupt change in pitch angle control and the curve 22 illustrates an actual corrective pitch angle control applied by the said rotor-cyclic pitch technique. The difference between the two curves is illustrated by 21.

[0115] The invention has been exemplified above with reference to specific examples of a wind turbine with a control system for controlling the wind turbine blades by pitch mechanisms. However, it should be understood that the invention is not limited to the particular examples described above but may be designed and altered in a multitude of varieties within the scope of the invention as specified in the claims e.g. in using other formulas and/or measuring data as a supplement.

[0116] In the drawings the following reference numbers refer to:

• [0117] 1. Wind turbine or wind turbine system
• [0118] 2. Wind turbine tower
• [0119] 3. Wind turbine nacelle
• [0120] 4. Low speed shaft
• [0121] 4a. Center point of the low speed shaft
• [0122] 5. Wind turbine rotor blade
• [0123] 6. Wind turbine rotor with at least one blade
• [0124] 7. Tip point of a wind turbine rotor blade
• [0125] 8. Centerline of the low speed shaft
• [0126] 8a. Formalized centerline of the low speed shaft
  • [0127] Vertical center line of the rotor blade through the center point of the low speed shaft
• [0128] Root of wind turbine rotor blade′
• [0129] Sensor means
• [0130] Computing means
• [0131] Pitch control system
• [0132] Example of a desired step pitch angle
• [0133] Example of an actual rotor-cyclic correction of the pitch angle
• [0134] Angle error—rotor-cyclic angle correction
• [0135] Idealized m_tilt
• [0136] Filtered m_tilt
• [0137] Idealized m_yaw
• [0138] Filtered m_yaw
• [0139] Angle error—harmonic angle correction
• [0140] Example of an actual harmonic correction of the pitch angle
• [0141] ψ. Azimuth angle for rotor blade 1 relative to a fixed vertical reference position