Stall Flag Diagnostics of the Aerpac 43M Rotor

ECN-C--01-042

STALL FLAG DIAGNOSTICS OF THE AERPAC 43M ROTOR

G.P. Corten

stall strip vortex generators initial situation with two problems:

1. overpower requires cut- out at 16 m/s 2. production below rated disappointing

with stall flag diagnostics

adaptations based on general knowledge added to increase power below rated removed, since it only reduced power below rated shifted since they caused the overpower

results: 1. overpower vanished 2. production below rated increased

added to reduce overpower added to increase power below rated

results: 1. more overpower 2. less production below rated

May 2001 Stall Flag Diagnostics of the Aerpac APX43 Rotor ECN-C--01-042

Stall Flag Diagnostics of the Aerpac 43m Rotor

G.P. Corten

2 Stall Flag Diagnostics of the Aerpac 43m Rotor

Preface

This project was carried out on request of the blade manufacturer Aerpac. Regretfully, Aerpac went bankrupt before the project was finished. Since most of the work had been completed, it was decided by ECN to finish the project by means of this brief report. During the project, ECN wrote 6 proposals for vortex generator patterns on Aerpac rotors. These proposals may contain confidential information of third parties (clients of Aerpac) and therefore they are collected in a separate confidential report: 'Appendices to Stall Flag Diagnostics of the Aerpac 43m Rotor'.

Distribution List

Aerpac: J. de Boer 1-5 H. Heerkes 6 R. Roelofs 7

H.J.M. Beurskens 8 E. Bot 9 G.P. Corten 10-15 F. Saris 16 J.G. Schepers 17 H. Snel 18 Central Archives ECN 19 Archives ECN Wind Energy 20-25

Confidentiality

This report remains confidential until 1 year after publication

Contents

Stall Flag Diagnostics of the Aerpac 43m Rotor 1 First Stall Flag Measurement 2 2 Second Stall Flag Measurement 6 3 Vortex Generator Modelling 9

1 Stall Flag Diagnostics of the Aerpac APX43 Rotor ECN-C--01-042

4 Conclusion and Main Results 11

2 Stall Flag Diagnostics of the Aerpac 43m Rotor Stall Flag Diagnostics of the Aerpac 43m Rotor

The Dutch blade manufacturer Aerpac requested the author in 800 40 the beginning of 1999 to 700 20 improve the stall behaviour of 600 0 their APX43 rotor. Figure 1 500 design -20 shows several power curves of 400 initial -40 the rotor. If we look at the 300 ECN -60 P [kW] intended curve ‘design’ and difference delta P [kW] measured curve ‘initial’, then 200 -80 we see the discrepancy at the 100 -100 start of the project. The turbine 0 -120 of 600 kW rated power could 6 8 10 12 14 16 18 20 22 24 maximally produce 660kW, wind speed [m/s] but, above 16 m/s, when the rotor captures even more than figure 1 The predicted power curve of the APX43 700 kW, the turbine will be ‘design’ (1.86 GWh), the measured curve before the halted. This is needed to project started ‘initial’ (1.65 GWh), and the measured prevent the turbine from curve after the first stall flag measurement ‘ECN’ (1.79 overloading or overheating, GWh). The scale on the right-hand side refers to the but the penalty is of course a difference between the two measured curves. production loss. We also see that the rotor produces less than predicted below rated and in particular at approximately 10 m/s. This again represents production losses. Both deviations caused a loss of approximately 10% relative to the design.

Our approach of these problems was to diagnose the stall behaviour with stall flags. Subsequently we will try to change the passive power control by adaptation of the stall behaviour, which we base on results of the stall flag diagnostics. Two series of stall flag measurements were carried out, one in July 1999 and one in March 2000. After the measurements both problems were solved by changing the vortex generator positions and by removing the stall strips.

1 First Stall Flag Measurement

During the first measurement for Aerpac we had periods of low wind and periods of rain, causing the flaps to stick to the blade surface. Only one measurement could be made. We present the most valuable result, namely for the APX43 blade in standard layout (see figure 2) with some adaptations. Delft University of Technology, who had an important input in the aerodynamic design of the APX43, proposed to install stall strips at approximately 0.85R to

3 Stall Flag Diagnostics of the Aerpac APX43 Rotor ECN-C--01-042 avoid overpower, and to extend the line of vortex generators from 0.52R to 0.58R to increase the power below rated. The power curve ‘initial’ in figure 1 refers to this layout. After installation of the stall flags 0 dg flags (as shown in figure 2), 1.5 vortex we started the turbine reference several times, while 1.0 flags 60 dg recording the stall flags signals responding to the 0.5 increasing rotation speed, [m]chord and thus the increasing tip 0.0 speed ratio λ. The video 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 frames were analysed by radial position r/R the image-processing program. The frames were figure 2 The pattern during the first measurement. binned on tip speed ratios The definition of the flag angle is given in figure 3. The and we obtained table 1 as vertical line at 0.86R is the tip division and the stall output. The table displays strip is located just to the left of it at the leading edge. the average signals of each The reference points are required for the automatic pair of adjacent trailing image processing edge stall flags. The second row shows the radial positions from the tip to the λ↓ # frames↓ strip vortex generators root of the blade. The first r/R → 1.9 .8 .8 .7 .7 .6 .6 .5 .5 .4 .4 .3 .2 .2 .1 column shows the value for 2.0 57.3.6 111111 .5 .8 .9 .8 .7 .6 .7 .4 2.3 109 .8 .9 .911111 .7 .9 .8 .9 .8 .9 .8 .3 λ for each bin and the 2.6 69.6.8 111111 .5 .9 .9 .8 .8 .9 .9 .4 second column shows the 2.9 48.9.9 111111 .7 1 .9 1 .8 .8 .9 .3 3.2 54 .8 1 111111 .6 1 1 1 1 .9 .9 .3 number of video frames in 3.5 38.5.8 1111.9.9 .5 1 1 1 .8 .8 1 .8 each bin. Therefore, the 3.9 25.4.9 11111.9 .5 .7 1 .9 .8 .7 1 .5 first row of data shows that 4.2 29 .6 1 11111.9 .2 .9 1 .9 .9 .9 1 .4 57 video frames were 4.5 28 .7 1 11111.5 .1 .4 .7 .6 .7 .7 1 .5 analysed for the λ=2.0 bin, 4.8 23 .5 1 11111.0 .1 .2 .3 .4 1 .5 5.1 17 .3 11111 .0 .2 .1 .1 1 .4 and the first two stall flags 5.4 21.1.4 1.5111 .2 .3 .2 .1 1 .7 on the tip were open 5.8 21.0.6 11111 .0 .3 .1 1 .6 (=visible) during a fraction 6.1 21 .4 1.9111 .0 1 .8 0.3 of the 57 frames. The 6.4 14 .0 1 .6.4.4 .1 1 .8 6.7 19 .2 1 .1.1.2 1.8 column with the heading 7.0 19 .3 1.2.6.9.9 .0 .1 1 .6 ‘strip’ shows the signal of 7.3 20 .0 .6 .0 .0 .0 1.7 the pair of stall flags behind 7.7 14 .2 1.8 the stall strip. We see that 8.0 22 .1 1.8 λ 8.3 22 .0 1.7 this strip causes stall for < 8.6 13 0 1.8 7.5. The tip speed of this 8.9 11 1.9 rotor is 61.5 m/s, so stall 9.2 186 1.8 9.5 151 1.9 occurred above 8 m/s, table 1 The rough signals of the stall flags averaged which is much too soon. over several starts. Each column presents the average According to the design signal of two adjacent stall flags. We see the advance of stall should occur at 13 - 14 stall caused by the stall strip already causing losses for m/s. The columns in the λ λ≈ λ≈ vortex generator area show =8, the vortex generators delay stall from 8 to 4.5. that these generators

4 Stall Flag Diagnostics of the Aerpac 43m Rotor

effectively delay stall. We wind turbine blade γ angle see that the switching over of a stall flag is gradual. The λ- values in the table are center of obtained from ΩR/U. For U rotation flap we used the average of the anemometer signal, obtained at hub height from a mast figure 3 Definition of the hinge angle. upwind, for R the known rotor dimension and for the angular frequency Ω the rather precise result of the 10 0.2-0.6c 0.40 image processing. Although the error in U makes λ 9 trailing edge 0.35 vortex somewhat uncertain, there is 8 0.30 also a systematic variation in 7 0.25 U due to wind shear. The 6 0.20 stall flags usually light up in 5 0.15 the upper half and close down in the lower half, lambda-stall 4 0.10 vg x/c position indicating that the wind 3 0.05 speed in the lower half is less 2 0.00 than that in the upper half. 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 The switching of the stall flags will take place at radial position r/R distinct angles of attack, which in our methodology λ figure 4 The -stall curves of several starts during the results in a range of λ- first measurement. The turbine reaches rated power for values. Using all data, we λ≈ 4.5. The surface between 0.45R and 0.55R is too late can however estimate the and may cause overpower. The entire outer blade ranging mean λ when each stall flag from 0.6R to 0.9R shows premature stall. In particular at switches over. For the stall the stall strip (0.85R) stalling is much too soon, which flags on the trailing edge reduces production below rated. those λ-values are given in figure 4. In the area below flow direction range in deep stall the curves, the flow is separated, above the curve it is still attached. Such λ-stall graphs display how stall extends over the rotor and this helps us to see which areas stall too late or too soon. In the design the rotor reaches rated power at approximately λ=4.5, so stall should not occur above this value. Since we flow direction range have overpower, stall is not sufficient for moderate stalling below it. We see that up to approximately 0.45R the behaviour is figure 5 The flow direction estimated from good. Moving outwards the vortex the stall flags under different hinge angles. generators between 0.45R and 0.55R

5 Stall Flag Diagnostics of the Aerpac APX43 Rotor ECN-C--01-042 are too effective; the blade remains attached below λ=4.5. The additional row of vortex generators at approximately 0.55R was meant to reduce overpower, but appears to be responsible for overpower. The stall flags placed under a hinge angle of 60° (defined in figure 3) correspond to the lower values of the zigzag line labelled ‘0.2c-0.6c’ in figure 4, so they close sooner during a start. This indicates that, with decreasing angle of attack, the separated flow over the blade changes direction from the reversed to the radial direction, see figure 5. This is as expected, since at large angles of attack (at low tip speed ratios) rotation has little effect on a reversed flow, which then behaves as a 2-dimensional separated flow. Between 0.6R and 0.9R the blade stalls much too soon, especially in the stall strip area. This clearly is the reason for the losses below rated. Another observation is that the stall strip has only effect on the stall flags behind it, so its influence is restricted to its physical dimensions. Finally, we see that the tip is a bit late in stalling. 0.50 We come to the remarkable conclusion that both measures 0.40 proposed to correct the stall behaviour made the problems 0.30 worse! Using stall flags, we 0.20 initial could clear up what really ECN 1 happened, why the corrections 0.10 ECN 2 didn’t help, and most vg-position x/c 0.00 importantly, what had to be done: Remove the stall strips, 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 since the blade stalls too soon in radial position r/R this range even without them. figure 6 The vortex generator patterns. Install vortex generators up to much larger radial positions, say 0.8R. Reduce the effectiveness of the vortex generators between 0.45R and 0.6R by moving them to a larger chord-wise position, and maybe, make the tip stall sooner. So we designed a new vortex generator pattern (see figure 6, 'ECN 1'). The power curve was measured again with this pattern and figure 1 shows the result. We see that the overpower problem has disappeared. While the turbine had to be halted above 15m/s in the initial situation, it could now operate up in its full design range. The difference of the two power curves (also shown in the figure) clarifies that the power below rated has improved, especially around 10 m/s where the rotor captures much energy. Due to both effects the production increased by 8%.

6 Stall Flag Diagnostics of the Aerpac 43m Rotor

2 Second Stall Flag Measurement

We were not satisfied with the above improvement below rated and decided to carry out a further measurement. When we returned to the turbine, we found 7 metres of vortex generators under the turbine. These rows had come loose and thus the power curve ‘ECN’ of figure 1 does not correspond precisely to the intended configuration. From the previous measurement we learned the stall flags should preferably be put immediately behind vg's. Only then, a good stall flag impression of the effect blade 1, unchanged of the vg is obtained. The vortex generators most interesting 2.0 reference marker configuration during the 1.5 second measurement is 1.0 shown in figure 7. The 0.5 pattern according to our

chord [m]chord 0.0 first advice was installed on blade 1, a new 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 configuration on blade 2 radial position [r/R] and the same stall flag pattern on blade 3, but blade 2, vg's >.55r/R at 0.39x/c, stall flag here without vortex vg's for .32

chord [m]chord obtain more accuracy. 0.0 Figure 8 shows the 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 trailing edge signals of radial position [r/R] the three blades. We immediately see the large blade 3, all vg's removed stall flag difference between blade vortex generators 3 (without vortex reference marker generators) and the two 2.0 stall strip -45 deg 1.5 other blades. Figure 9 1.0 shows the results for the 0.5 stall flags behind the vortex generators near chord [m]chord 0.0 the leading edge. As 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 above, we used these radial position [r/R] signals to design the vortex generator pattern figure 7 The pattern of the second measurement. indicated with ‘ECN 2’ Blade 1 has the pattern according to the first advice of in figure 6. This advice is ECN. Blade 2 has a second new pattern. On blade 3 the been protected by ECN's vortex generators were removed, while the blade has the patent application same stall flag pattern as blade 2, except of the smaller 1012949. At this moment number of stall flags on the trailing edge. (September 2000), the

7 Stall Flag Diagnostics of the Aerpac APX43 Rotor ECN-C--01-042 vortex generator pattern has just been installed, but the start2&3, trailing edge blade 1 power curve is not yet blade 2 known. This did not prevent 11 blade 3 us to apply a comparable 10 pattern already to a larger 9 Aerpac rotor, the APX45. In 8 this case we could certify 7 (see figure 10) that the power 6 rises well up to rated value lambda-stall and that the final level is flat, 5 although we do not have data 4 above 18 m/s. This power 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 curve is excellent for a stall radial position r/R turbine. figure 8 Trailing edge λ-stall graphs for the The following arguments, second measurement. Blades 1 and 2 show small derived from the second differences, while blade 3 without vortex generators measurement, have been the attaches much later. reason for a second ECN- advice for the vortex generator positions. In figure 8 we see that blade 1 remains attached up to λ≈4 between 0.4R and 0.5R, while blade 2 start2&3, behind vg's blade 1 blade 2 switches to the stalled state 10 blade 3 around λ≈5. The turbine 9 reaches nominal power for 8 λ≈4.5, and attached flow 7 below this value may cause 6 overpower. If the blade 5

without vortex generators lambda-stall 4 would also have the same 3 pattern (which will become 2 the situation of practice), the 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 induction will be more and the angles of attack thus radial position r/R smaller (as if λ is higher). figure 9 Leading edge λ-stall graphs. The Therefore, the position of the differences at the tip are due to stall strips between vg’s on blade 2 (giving stall 0.92R and R on blades 2 and 3. below λ≈5 in the measurement, but a bit less in practice) is acceptable. The stall flags at 0.7R and 0.73R on blade 1 were out of order, therefore the line is not continued. However, we can see in figure 9 that blade 1 stalls for λ≈6 between 0.55R and 0.65R and at λ≈5 for blade 2. So, the position on blade 2 (0.39c) is fine again. Moving outboards to 0.8R the stall tip speed ratio become slightly higher (from about 5 at 0.6R to about 6 at 0.8R), indicating that the vortex generator can be moved a bit towards the leading edge so that they become more effective.

8 Stall Flag Diagnostics of the Aerpac 43m Rotor

We finally see that the tips 800 stall rather late. This might be a reason of overpower, so we may need stall strips on 600 the tips. 400 P-design measurement Pe [kW] Pe 200

0 5 10152025 wind speed [m/s]

figure 10 APX48 power curve, measured and predicted for a vortex generator pattern based on the APX43 experience.

9 Stall Flag Diagnostics of the Aerpac APX43 Rotor ECN-C--01-042

3 Vortex Generator Modelling

Vortex generators are a powerful tool to correct even large deviations in the stall behaviour, but we do not know how their efficiency depends on the radial and chord-wise position. Let us derive their effect from the stall flag observations. We start with expressing the stall behaviour in terms of stall-angles instead of stall tip speed ratios. Remember that the induction has to be built up and that it thus will be too low during a start. Blade – element - momentum theory is based on a steady wake or on steady induction, so the estimated angles will be systematically too low during a start. By calculating the angles of attack with and without induction, we find a maximum and a minimum start2&3, trailing edge, no induction blade 1 value for the angles. This error blade 2 25 interval is not precisely the blade 3 same for the three blades, 20 ind-error since they stall at different instants, but the values are 15 almost equal. Therefore we determined this average error 10 and included it in figure 11 5 and 12 with stall-angles for the [deg] AOA three blades calculated without 0 induction. The absolute error 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 due to induction is radial position r/R approximately half the induction error in the figure, so figure 11 The stall angles at the trailing edge. The ° it is approximately +/-1 at the maximum induction error is included. The angles will tip, increasing to +/-3° at 0.4R be too high, since 'no induction' was assumed. and +/-6° at 0.2R. If we record stall patterns during steady turbine operation, the induction is also steady and the start2&3, behind vg's, no induction blade 1 error vanishes. By comparing blade 2 the stall patterns obtained 30 blade 3 during steady operation with 25 ind-error those during a start, we can estimate the induction error 20 and correct for it. Even without 15 the steady measurements, we 10 can compare different blades with different configurations. [deg] AOA 5 The differences will depend 0 much less on the induction, 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 since the error is approximately equal for all radial position r/R blades. We calculated the stall figure 12 The stall flag angles at the leading edge, delay angles due to the vortex again too high as in figure 9. generators for blades 1 and 2

10 Stall Flag Diagnostics of the Aerpac 43m Rotor relative to blade 3 and plotted 10 the results in figure 13. The scatter in the points is 8 representing the statistical error of approximately +/-0.5°, being 6 increased by the subtraction. The figure shows how effective 4 vortex generators are in terms of blade 1, trail. edge 2 degrees of stall delay. So stall blade 2, trail. edge flags allow us to measure the [deg] angle delay stall 0 stall delay and effective range of vortex generators as a function 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 of their radial and chord-wise radial position [r/R] positions, with approximately +/-0.5° accuracy. figure 13 Stall delay of blades 1 and 2 with vortex generators relative to blade 3 without.

11 Stall Flag Diagnostics of the Aerpac APX43 Rotor ECN-C--01-042

4 Conclusion and Principal Results

We come to the remarkable conclusion that both measures (based on general knowledge) to correct the stall behaviour made the problems worse! Using stall flags, we could clear up what really happened, why the corrections didn’t help, and most importantly, what had to be done: Remove the stall strips, since the blade stalled too soon. Install vortex generators up to much larger radial positions, say 0.8R. Reduce the effectiveness of the vortex generators between 0.45R and 0.6R by moving them to a larger chord-wise position, and maybe, make the tip stall sooner. The adaptations are shown in figure 14. As a result the production increased by 8% and the overpower vanished. In a second stall flag measurement further optimisation was intended. However before an even better vortex generator pattern could be installed the blade manufacturer Aerpac went bankrupt.