Evaluation of a wind power parameterization using tower observations
Lazarus, Steven M.
Bewley, Jennifer L.
MetadataShow full item record
The spatial and temporal components of a published wind power parameterization method are evaluated using observed winds (9 m to 90 m) from 7 years of data collected at four towers in the Kennedy Space Center/Cape Canaveral Air Force Station network. The temporal component is governed by two parameterization inputs which represent the amplitude and mean of an assumed sinusoidal diurnal variation of the ratio of the 80 m to 10 m winds, respectively. Comparison with tower observations shows that the estimates of the mean ratio are robust but biased high, indicating that the temporal variation of the observations can be approximated by, but is not, a pure sinusoid. The observed and parameterized amplitude are poorly correlated as the amplitude estimate is sensitive to small phase shifts in the diurnal variation of the ratio of the wind speeds. The observed annual wind power for the site is consistent with what is known about the wind energy potential in Florida, while wind power estimates based on the temporal extrapolation are more than twice that observed. The erroneous wind power estimate is shown to be related to an assumption that negative amplitude estimates are indicative of a diurnal phase shift in the ratio of the 80 m to 10 m winds. When the assumed phase shift is removed from the data and the observed inputs are applied, the parameterized 80 m wind power estimates were comparable to those observed. The spatial component of the parameterization depends on a least squares fit of four different wind extrapolation methods to 0000 and 1200 UTC wind profiles. Comparison of the observed 80 m winds with those obtained via the least squares method indicates that the 0000 UTC 80 m wind and power estimates exceed (by as much as 1 m s -1 and 200 W m-2, respectively) the observed 80 m winds and power for all months. When compared against the 0000 UTC regression, the 1200 UTC 80 m wind and power estimates are closer to the observed power for all months. The power law often yields the lowest residual (on the order of 15-20% of the time) but is also largely responsible for high power estimates. The positive power bias (i.e., estimated greater than observed) is a result of the combination of two factors, namely, (1) the 0000/1200 UTC profiles and 10 m winds are, in general, not representative of the daily averaged values, with the 0000 and 1200 UTC 10 m winds less than their daily average, and (2) the differences between the observed 80 m and 10m 0000 and 1200 UTC winds are greater than the difference between the daily averaged 80 m and 10m mean winds. These results illustrate the potentially problematic nature of combining the lowest residual producing (extrapolation) method, obtained from a fit to the 0000 or 1200 UTC wind profiles, and the daily-averaged 10m wind to produce a daily-averaged 80 m wind estimate. Copyright 2005 by the American Geophysical Union.