A novel approach to statistical‐dynamical downscaling for long‐term wind resource predictions

Abstract

ABSTRACTA new method for the long‐term prediction of the wind resource based on the concept of statistical‐dynamical downscaling is presented. This new approach uses mean sea level pressure maps from global reanalysis data (National Centers for Environmental Prediction Department of Energy Atmospheric Model Intercomparison Project (NCEP‐DOE AMIP‐II)) and image processing techniques to identify a synthetic reference period which optimally matches the corresponding long‐term maps. Four different image processing techniques, averaged into one image similarity error index, are used to evaluate image similarity. A representative set of days is selected by requiring the error index to be minimal. Validation of representativeness in terms of the wind resource for the Iberian domain is performed against 10 years of measured wind data from Navarra (Spain), as well as mesoscale simulations of the Iberian Peninsula. The new approach is shown to outperform not only the industry‐standard method but also other recently proposed methods in its capability to achieve mesoscale level representativeness. A particular advantage of the new method is its capability of simultaneously providing a representative period for all potential wind farm sites located within large regional domains without requiring re‐running of the method for different candidate sites.