Method to Improve Offshore Wind Energy Resource Assessments Using Cokriging

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Monday, 18 January 2010: 1:45 PM
B305 (GWCC)
Michael J. Dvorak, Stanford University, Stanford, CA; and A. Boucher and M. Z. Jacobson

Wind energy represents the fastest growing renewable energy resource, sustaining double digit growth for the past 10 years with approximately 94,000 MW installed by the end of 2007. Although winds over the ocean are generally stronger and often located closer to large urban electric load centers, offshore wind turbines represent about 1% of installed capacity. In order to evaluate the economic potential of an offshore wind resource, wind resource assessments typically involve running large mesoscale model simulations, validated with sparse in-situ meteorological station data. These simulations are computationally expensive limiting their temporal coverage. Although a wealth of other wind data does exist (e.g. QuickSCAT satellite, SAR satellite, radar/SODAR wind profiler, and radiosounde) these data are often ignored or interpolated trivially because of the widely varying spatial and temporal resolution. A spatio-temporal cokriging approach with non-parametric covariances was developed to interpolate these empirical data and compare it with previously validated surface winds output by the PSU/NCAR MM5 for coastal California. The spatio-temporal covariance model is assumed to be the product of a spatial and a temporal covariance component. The temporal covariance is derived from in-situ wind speed measurements at 10 minutes intervals measured by offshore buoys and variograms are calculated non-parametrically using a FFT. Spatial covariance tables are created using MM5 or QuikSCAT data with a similar 2D FFT method. The cokriging system was initially validated by predicting “missing” hours of PSU/NCAR MM5 data and has displayed reasonable skill. QuikSCAT satellite winds were also substituted for MM5 data when calculating the spatial covariance, with the goal of reducing the computer time needed to accurately predict a wind energy resource.