An hybrid approach for downscaling RAMS data for wind resource assessment in complex terrains

Estimating the wind resource distribution on complex terrains can be very difficult because of the complex interaction of meteorological and aerodynamics phenomena. For this reason the wind characterization of large areas needs the collection of field data in many different positions during a period covering all wind climatology seasonal variations. Meteorological models can be very useful to speed up the characterization of very large areas in terms of mean annual speed intensity and direction; unfortunately such models generally work on coarse grids and results for wind climatology are reliable only at very high levels above the ground. In the present work a new method to downscale climatology data elaborated by the meteorological models RAMS and WRF was developed and tested for the case of the M. Ginezzo wind site (ITALY). RAMS and WRF data was provided by La.M.M.A. and the anemometric data used for the method validation were provided and elaborated by Sorgenia S.p.a. Different approaches were used for downscaling annual wind speed and direction time histories estimated by the meteorological model: using CFD wind field distributions to evaluate speed-up and flow distortions or using a neural networks trained on a period of twenty days. The first technique (CFD) is generally more difficult to be tuned but it is able to give good results on large calculation domains while the second one (neural network) can give reliable results on a fast way generally only for a restricted area.