This preview shows page 1. Sign up to view the full content.
Unformatted text preview: eviations occur with positive 9 yaw error, where the tower shadow and yaw error effects on α
are complementary in φ.
At 80% span (Fig. 13), the changes in CN with V∞ and
ϕ are less pronounced than those at 30% span. However,
average CN values of 1.5 still far exceed quasi-static wind tunnel
performance data. Deviation in normal force is also less
prominent than for the 30% span location, but still of significant
magnitude. As at 30%, the standard deviation distribution is
asymmetrical with respect to yaw error, with the maximum
occurring at 25° yaw error.
CN mean values respond differently for the two blade
geometries. For the untwisted blade, the entire CN response
centroid is shifted to higher wind speeds at increasing span
locations (not shown). This effect appears to simply reflect the
decrease in α with increasing span for any wind speed.
Equivalent angles of attack are achieved with higher wind
speeds. A strong similarity exists in both cycle averaged and
dynamic response at incremental span locations of 47% and
63% (not shown) with peak magnitudes diminishing slightly
(≈10%) at 80% span. In contrast, CN mean values for the
twisted blade do not shift. At low wind speeds, CN contours are
quite similar across all span locations, reflecting the uniformity in αi at the designed wind speed. Increasing wind speed
broadens the bands at any span location but does not shift the
overall response curve. Again, this is consistent with the nonuniform increase in α with higher wind speeds over the span.
Dynamic activity, as measured by the CN standard
deviation, decreases with increasing span for both geometries.
However, twisted blade dynamic activity is much greater. Blade
root CM data obtained from strain gage measurements are
plotted as CM means and standard deviations for both the
untwisted and twisted blade geometries (Figs. 14 and 15).
Standard deviation of root flap bending is greatest at increased
V∞ for both positive and negative values of ϕ. This unsteady
behavior correlates best with the high incidence of cycles where
Cp < -8.0 in Fig. 8. Logically, the high percentage of cycles
with large pressure fluctuations at 80% span will produce the
greatest standard deviation in the root flap load. Interestingly,
however, 80% span for the untwisted blade shows similar
pressure cycle activity without the resulting CM behavior.
Whether this effect reflects data density variance, variance due
to quasi-static mean CN participation at various spans, or a true
three-dimensional dynamic effect resulting from geometry or
separation effects has not been resolved at this time. Figure 14: Average and standard deviation for CM for untwisted
blade. Figure 15: Average and standard deviation for CM for twisted
blade. 10 CONCLUSIONS
Aerodynamic performance was compared for two
rectangular planform wind turbine blades having S809 airfoil
cross-sections. These blades differed only in three-dimensional
geometry, with one blade being untwisted and the other having
an optimized twist distribution. These comparisons were based
on field test data acquired at the National Wind Technology
Center. Both blades exhibited two-dimensional aerodynamic
performance at inflow conditions where the local blade angle
was below static stall.
Significant variations in performance were noted under
near- and post-stall conditions. Peak pressure distributions and
mean aerodynamic loads were far in excess of two-dimensional
wind tunnel data. Individual pressure time series for select test
cycles indicated both three-dimensional flow effects as well as
dynamic stall contributed to these enhanced loads.
Both blades were more dynamically active under nearand post-stall operation. This effect increased with both mean
velocity and yaw error. Certain combinations of V∞ and ϕ
exacerbated the effect and were correlated to cycles with high
peak pressure coefficients at 80% span. The twisted blade was
observed to be more dynamically active than the rectangular
blade. It is not clear if this effect reflects data density variance,
variance as a r...
View Full Document
- Spring '11