What can reanalysis data tell us about wind power?
Renewable Energy, (November 2015)DOI: 10.1016/j.renene.2015.05.027
Abstract
We present the first model of bias and uncertainty of reanalysis wind speeds. We use the model to simulate long-term wind generation in the central U.S. Annual wind energy variability at individual sites is 5–12\%. Annual wind energy variability for all sites in the central U.S. is 3\%. Portfolios of 2 wind farms reduce quarterly cash-flow variability by 1/3 vs. 1 farm. Reanalysis data sets have become a popular data source for large-scale wind power analyses because they cover large areas and long time spans, but those data are uncertain representations of ” true” wind speeds. In this work we develop a model that systematically quantifies the uncertainties across many sites and corrects for biases of the reanalysis data. We apply this model to 32 years of reanalysis data for 1002 plausible wind-plant sites in the U.S. Great Plains to estimate variability of wind energy generation and the smoothing effect of aggregating distant wind plants. We find the coefficient of variation (COV) of annual energy generation of individual wind plants in the Great Plains is 5–12\%, but the COV of all those plants aggregated together is 3.0\%. The year-to-year variability (of interest to system planners) shows a maximum step change of ∼10\%, and the wind power varies by ±7.5\% over a 32-year period. Similarly, the average variability of quarterly cash flow to equity investors in a single wind plant is 29\%, but that can be reduced to 18–21\% by creating small portfolios of two wind plants selected from regions with low correlations of wind speed.