Optimization of power production in a wind turbine at below rated power

Abstract

The rotor blades of a wind turbine each have a plurality of fiber-optic pressure variation sensors which can detect the onset of a stall condition. The output of the stall condition sensors is input to a stall count circuit which increases a stall count signal each time a stall indication is received. The stall count signal is decayed exponentially over time and the current signal is summed with the decayed signal from a previous sampling period to form a value from which a stall margin is determined. An : curve of tip speed to wind speed ratio against pitch angle reference is then determined from the stall margin.

Claims

1. A method of operating a wind turbine at below rated power, comprising: sensing stall conditions using one or more stall condition sensors on wind turbine rotor blades; determining a stall margin based on the sensed stall conditions, wherein the determination of the stall margin comprises determining a stall event count using a stall event counter based on outputs from the one or more stall condition sensors, the stall margin being determined from the stall event count, wherein the stall event count is dependent on the time between detected stall conditions output by the at least one stall condition sensor; selecting a curve of tip to wind speed ratio against blade pitch angle according to the determined stall margin; and varying the blade pitch angle of at least one of the wind turbine blades in accordance with the selected curve.

2. A method according to claim 1, wherein the stall event counter increases the stall event count on detection of a stall condition by the at least one stall condition sensor, and decays the stall event count over time.

3. A method according to claim 2, wherein an output from the at least one stall condition sensor is processed to determine an increase in a stall event signal, and scaled by an amount determined according to wind conditions.

4. A method according to claim 3, wherein the processed and scaled output is summed with a processed and scaled output from a previous sampling time period weighted by an amount determined according to wind conditions.

5. A method according to claim 4, wherein the summed outputs are mapped to a stall margin and the curve of tip speed to wind speed ratio against blade pitch angle determined from the mapped stall margin.

6. A method according to claim 5, wherein the stall margin is varied between a nominal value and an optimal value depending on the stall event count output.

7. A method according to claim 3, wherein a first constant is determined on the basis of wind speed and wind turbulence, and the first constant is used to scale the increase in the stall event signal.

8. A method according to claim 1, wherein the at least one stall condition sensor comprises a plurality of stall condition sensors on each of the rotor blades of the wind turbine.

9. A method according to claim 8, wherein the stall condition sensors are fibre-optic pressure variation sensors.

10. A method according to claim 9, wherein the fibre-optic pressure variation sensors comprise a membrane arranged over an aperture in a blade surface and a fibre-optic light transmitter and a fibre-optic light receiver arranged within the aperture to detect vibration of the membrane by detecting fluctuations in the intensity of light received by the fibre-optic light receiver.

11. A method according to claim 1, wherein the varying of the blade pitch angle of at least one of the wind turbine blades in accordance with the selected curve varies the pitch angle of all the blades by a common amount.

12. A control system for operating a wind turbine at below rated power, comprising: one or more stall condition sensors for sensing stall conditions on wind turbine rotor blades; a stall margin determination arrangement for determining a stall margin based on sensed stall conditions, wherein the arrangement for determining the stall margin comprises a stall event counter for determining a stall event count based on outputs from the one or more stall condition sensors, the stall margin being determined from the stall event count and the stall event count being dependent on the time between detected stall conditions output by the at least one stall condition sensor; a controller for selecting a curve of tip to wind speed ratio against rotor blade pitch angle according to the determined stall margin; and a pitch angle controller for varying the blade pitch angle of at least one of the rotor blades in accordance with the selected curve.

13. A control system according to claim 12, wherein the stall event counter increases the stall event count on detection of a stall condition by the at least one stall condition sensor, and decays the stall event count over time.

14. A control system according to claim 13, wherein the stall event counter determines an increase in a stall event signal from the at least one stall condition sensor, and the stall event signal is scaled according to wind conditions.

15. A control system according to claim 14, comprising a summer for summing the scaled stall event signal with a scaled stall event signal from a previous sampling time period weighted by an amount determined according to wind conditions.

16. A control system according to claim 15, comprising a device storing a map of stall margins and corresponding curves of tip speed to wind speed ratio against blade pitch angle.

17. A control system according to claim 16, wherein the stall margin is varied between a nominal value and an optimal value depending on the stall event count.

18. A control system according to claim 14, wherein a first constant is determined on the basis of mean wind speed and wind turbulence, and the first constant is used to scale the increase in the stall event signal.

19. A control system according to claim 12, wherein the at least one stall sensor comprises a plurality of stall condition sensors on each of the rotor blades of the wind turbine.

20. A control system according to claim 19, wherein the stall condition sensors are fibre-optic pressure variation sensors.

21. A control system according to claim 20, wherein the fibre-optic pressure variation sensors comprise a membrane arranged over an aperture in a rotor blade surface and a fibre-optic light transmitter and a fibre-optic light receiver arranged within the aperture to detect vibration of the membrane by detecting fluctuations in the intensity of light received by the fibre-optic light receiver.

22. A control system according to claim 12, wherein the pitch angle controller is a common pitch angle controller for varying of the blade pitch angle of the wind turbine blades in accordance with the selected curve by a common amount.

23. A wind turbine having a control system according to claim 12.

Description

(1) Embodiments of the invention will now be described, by way of example only, and with reference to the accompanying drawings, in which:

(2) FIG. 1, (referred to above) shows a typical : curve for a wind turbine below rated power;

(3) FIG. 2, (referred to above) shows a lift curve for a typical wind turbine blade;

(4) FIG. 3a and FIG. 3b show a fibre-optic pressure variation gauge for detecting blade stall;

(5) FIG. 4 shows, schematically, an embodiment of the invention which estimates a stall margin;

(6) FIG. 5 is a block diagram showing the stall margin estimator of FIG. 4;

(7) FIG. 6 illustrates how counting stall events may be used to adjust a pitch reference;

(8) FIG. 7 shows how stall event counter constants may be adaptively adjusted in dependence upon wind turbulence and mean wind speed;

(9) FIG. 8 is a graph of stall events against stall margin; and

(10) FIG. 9 shows how a stall event count may be mapped to a stall margin and converted to a collective blade pitch reference.

(11) In the embodiments of the invention to be described, the stall margin is adaptively controlled in response to sensed wind conditions. However, for this to be possible it is necessary to be able to determine accurately the likelihood of stall occurring. FIG. 3 illustrates a stall sensor which may be used for this purpose. It is to be understood that this sensor is exemplary and that other stall sensors, either direct or indirect, may be used. The stall sensor of FIG. 3 is described firstly in WO-A-2011/015383 the contents of which are incorporated herein by reference. FIG. 3 shows one embodiment of the disclosure of WO-A-2011/015383. FIG. 3a shows a wind turbine blade 100 in cross section with air flowing over the aerodynamic surface from the leading edge 110 to the trailing edge 120. Initially, the flow will be laminar, but at some point along the upper surface the laminar flow will detach and the flow will become turbulent, potentially leading to a stall. The point at which the flow transitions from laminar to turbulent will depend on many factors including angle of attack, pitch angle, wind speed, condition of the air and the condition of the blade surfaces.

(12) A plurality of turbulence sensors 130 are arranged along the upper surface of the blade over the rear half of the blade approaching the trailing edge. As described fully in WO-A-2011/015383, the sensors are placed at points along the blade where the boundary layer is likely to separate and a stall initiated. FIG. 3b shows one of the turbulence sensors in more detail. The sensor comprises a pair of fibre-optic devices 200, 210. The first device comprises a fibre-optic cable 220 and a light transmitter 230 and the second device comprises a fibre-optic cable 240 and a light receiver 250. The two devices are arranged within a small aperture in the blade surface. Due to the size of the fibre-optic cables, this aperture may be as small as 5 mm diameter or less. The fibre-optic cables extend into the interior of the blades with the light transmitter 230 and light receiver 240 positioned at the edge of the aperture. A flexible or resilient membrane 260 is placed across the aperture supported above the aperture as shown in FIG. 3b. The membrane is sensitive to pressure variations caused by turbulent air flow across the upper surface of the blade and will vibrate both depending on the wind speed and the degree of turbulence. The light transmitter 230 and light receiver 240 are positioned such that light from the transmitter 230 is reflected off the inner surface of the membrane 260 to the receiver 240. The light received by the receiver will depend on the vibrations of the membrane and constructive and destructive interference between the transmitted and received light will lead to fluctuations in light intensity at the receiver which can be monitored to determine the level of vibration of the membrane and, from that, can be calibrated to measure the turbulence of the blade surface and function as a stall detector. As multiple sensors are used at key points of the blade, the collective signals from all these sensors can be used to detect the onset of a stall and the development of a stall over the blade surface.

(13) In FIG. 3, a plurality of stall sensors are arranged along the cross section of the blade, on the upper blade surface. As explained below, these sensors are used as the basis of an adaptive control algorithm of the stall margin which, in turn, enables a more aggressive : curve to be derived and so enables greater power production below rated power conditions.

(14) We have appreciated that under benign wind conditions where wind turbulence is low and wind direction is relatively constant, a smaller stall margin is acceptable without compromising emitted acoustic noise. That is, under those conditions, there is a lower likelihood of stalls commencing leading to the emission of noise. Thus, under these conditions, the turbine can be run with a more aggressive : curve. Where conditions are less benign with a more turbulent airflow and more variation in wind direction, the : curve is less aggressive.

(15) FIG. 4 shows, schematically, how a stall margin estimator is used to provide an input to an : curve calculator 300. In FIG. 4, each of the turbine rotor blades 310 A, B and C has an array of stall sensors as described with respect to FIG. 3 above. The outputs of these stall sensors are provided to a stall margin estimator 320 which analyses the stall sensor outputs and determines the appropriate stall margin for the conditions sensed and then outputs a stall margin signal to the : calculator 300.

(16) The : calculator 300 receives the tip/wind speed ratio as its other input and calculates the appropriate collective pitch reference signal as its output.

(17) In the arrangement in FIG. 4, the stall margin estimator 320 estimates the margin of stall based on the signals from the multiple stall detectors on each of the blades.

(18) When a high stall margin is estimated, the : curve at 300 is adjusted aggressively towards an optimum curve which can maximise power output. However, when a low stall margin is estimated, the adjustment is more conservative towards a nominal curve. Thus, in practice, the blades are pitched more into the wind when a high stall margin is estimated, indicating a lower likelihood of stall and benign conditions, than under normal conditions when a lower stall margin is estimated. Thus, under the high stall margin conditions, more power can be extracted from the wind than under normal operating conditions.

(19) FIG. 5 shows the stall estimator in greater detail. It is to be understood that this is only an example of how the stall estimator has been constructed and other variations are possible and will occur to those skilled in the art.

(20) In FIG. 5, the input s is the output from the stall sensor described in FIG. 3. This may be multiple inputs, one from each sensor or a simple input indicative of a sensed stall by any of the sensors. In this embodiment the input s of the signal processing block 405 is binary and will be zero unless any of the sensors detect a stall in which case the input will be 1. Other analog or digital input configurations could be used. Although FIG. 4 shows an array of sensors on each blade, a single sensor on each blade could be used. For each of these there is an instance of the Stall Event Counter so that the stall margin is estimated for each blade. In one example, the minimum of the three stall margin estimates is used to adapt the curve, and therefore the collective pitch angle in the manner described below. In another example, three instances of the : curve may be determined, one for each blade, each curve being based on a single sensor or an array of two or more sensors. Each instance of the curve may be used to set an individual pitch angle for a blade instead of a collective pitch angle and the control signal applied to the pitch controller is therefore an individual blade pitch control signal for each blade.

(21) When a stall event is detected, by the stall sensor 400, a stall event counter 410, which has a value c having an exponential decay is increased by a constant k.sub.1 and then reduced towards zero at a rate dependent on a second constant k.sub.2. Thus, if several stall events are detected over a relatively short period of time, for example from several sensors on different blades, the value of will increase drastically. If the period between stall events is relatively long, c will remain relatively low.

(22) The stall event counter in FIG. 5 is a discrete-time implementation in which T.sub.s is the sample time, Z.sup.1 is one sample delay block and U>U.sub.z.sup.1 is a logic operator detecting an increase in the sensor signal. Thus, an output from the stall sensor 400 provides the input s to logic operator 420 which determines on the basis of the input from the sensor, whether the number of stall events is greater than in the previous time period, for example, by incrementing a counter. The logic operator produces an output of which is multiplied by a constant at amplifier 430, the output of which provides an input to a summer 440. The output of the summer is input into a time delay 450 equal to one sample delay and the delayed output provides the stall event counter signal c from which the stall width is calculated at block 460. The output of the delay 450 also form the input to an amplifier 470 which multiplies the delayed value by an amount (1e.sup.Ts/Kz) to provide a second input to the summer 440. This amplifier provides the exponential decay.

(23) The effect of the stall event counter can be seen in FIG. 6. In this figure, graph a) shows the output S of the stall counter against the number of sample times n; graph b) shows the output of logic operator U>U.sub.z.sup.1 420 over this period; and graph c) shows the stall event count signal . Graph a) shows that the input s to the counter has gone high in three separate time periods which produces a high output v from the logic operator 420. In Figure c) the magnitude of the increase in the stall event count signal for each detected stall event is equal to the first constant k.sub.1. The value of then decays exponentially until the next high output from the logic operator 420. In this example the second and third events are close together and so the value of after the second event has only decayed to about half its value at the time of the third event so that the value of rises above a threshold value .sub.0 which it remains above for some time until it decays below .sub.0 in the absence of a further event.

(24) The values of k.sub.1, k.sub.2 are dependent on wind conditions and can be adapted continuously to increase or decrease the weight given to the detection of stall events to the stall event counter. This may be done in correspondence with measured emitted acoustic noise and/or measured loads due to stall events. FIG. 7 shows how K.sub.1 may be adapted on the basis of emitted acoustic noise. FIG. 7 is a graph of mean wind speed against turbulence. Mean wind speed on the x axis has three critical values:cut-in where the speed is high enough for the rotor to start turning and generating power; rated, at which the turbine outputs its rated power value; and cut-out, where the wind speed is too high for safe operation or the turbine operation is shut down, for example, by feathering the blades or yawing the rotor out of the wind.

(25) In FIG. 7, six different regions are defined and the tuning considerations are as follows:

(26) TABLE-US-00001 Region 1 Wind speed above cut-in but The risk of stall events is low and the well below rated wind speed. influence of stall events or emitted noise is Turbulence is low. low, k.sub.1 is therefore set relatively low. Region 2 The wind speed is well The risk of stall events is low but the above cut-in but below influence of stall events on emitted rated. Turbulence is low. acoustic noise is high. k.sub.1 is therefore set in an intermediate region. Region 3 Mean wind speed is above The risk of stall event is low and the rated, but below cut-out, influence of stall on acoustic noise is low. and turbulence is low. k.sub.1 is therefore set low. Region 4 The mean wind speed is The risk of stall events is high but those above cut-in but well below events have a low influence on acoustic rated. Turbulence is high. noise. k.sub.1 is set to an intermediate region. Region 5 The mean wind speed is Both the risk of stall events and the well above cut-in, but below influence of those events on acoustic noise rated. Turbulence is high. levels is high. k.sub.1 is therefore set relatively high. Region 6 The mean wind speed The risk of stall events is high but those is above rated, but events have a low influence on acoustic below cut-out and noise. k.sub.1 is set to an intermediate region. turbulence is high.

(27) As can be seen from FIG. 6, the stall event counter has three ranges of values: below .sub.0; between .sub.0 and .sub.1; and above .sub.1. The value of is mapped to the stall margins by mapping function 460 in FIG. 5. An example of a suitable mapping is illustrated in FIG. 8 from which it can be seen that:

(28) where <.sub.0, the stall counter is relatively low and the optimum : curve may be used to optimise power production;

(29) where .sub.0.sub.1, the stall counter is in an intermediate region in which an : curve between the optimum and a nominal curve should be used to find a trade-off between optimal power production and indication of emitted acoustic noise;

(30) where >.sub.1, the stall counter is relatively high and the nominal : curve should be used to limit the emitted acoustic noise.

(31) FIG. 9 shows an example of how the value of in FIG. 9a may be mapped to the stall margin in FIG. 9b and the common blade pitch reference in FIG. 9c. The graph of stall rate count in FIG. 9a is that of FIG. 6c repeated for ease of comparison. Thus, it can be seen that the stall margin and the pitch reference are maintained at optimal except for a momentary excursion when the second event caused the value of to exceed .sub.0. However, as this decays away the optimal values are resumed until the third event where the value of exceeds .sub.0 but remains above .sub.0 while it decays. This causes a drop in towards nominal and a gradual recovery towards optimal. At this point the pitch reference increases from the optimal value towards the nominal value and then returns towards the optimal value following the defined mapping of stall margin to pitch reference for a given , referred to in FIG. 4 as the : adjustment. The adjustment in pitch reference may be made continuously, without discrete steps, by introducing a rate limiter on the stall margin signal. The mapping may be implemented as a look up table stored in memory within the turbine controller.

(32) The embodiment described may be implemented within a wind turbine controller or even at a higher level on a wind park controller that controls more than one wind turbine. Existing wind turbine controllers include a component which calculates the optimum collective pitch angle depending on the relationship between pitch angle and rotor speed. The embodiment described may conveniently be implemented as a modification of the existing pitch angle control system.

(33) Thus, by accurate measurement of stall events using stall sensors at suitable locations on the upper surface of the blades, the stall margin applied to the curve of pitch reference against tip/wind speed ratio can be controlled such that the power output from the turbine can be optimised when the risk of stall is low and the stall margin can be adaptively controlled according to the risk of stall and the effect of stall on a parameter such as emitting acoustic noise, or controlling blade loads, which are desirable to control.

(34) As well as enabling optimisation of energy generation at below rated wind speed, embodiments of the invention can also reduce maintenance costs as blades are prevented from operating continuously in the stall region. Moreover, the use of fibre-optic sensors, as described with respect to FIG. 3, while not essential, has the advantage that they are low cost and reliable which is highly desirable in a wind turbine which may be located in an inaccessible location.

(35) Many modifications to the embodiments described are possible and will occur to those skilled in the art without departing from the scope of the invention which is defined by the following claims.