As part of a PIER PIER1 and EPRI EPRI2 co-sponsored project an evaluation of the performance ofPrediktor at two California wind farms was made. This paper highlights the successes andshortcomings of the method and attempts to identify the sources of error. The successes are thatthe Predikor results can out-perform persistence forecasts within 6 hours. The shortcomings arethat in some months Prediktor forecasts can be too persistent and miss some high wind events.This is partly due to overly rigid model output statistics (MOS) corrections and in one case apoorly described diurnal cycle. The MOS corrections are fitted for the prevailing conditions fortwo 3-month periods in summer or winter. In connection with this there is a tendency for thepredictions to be less accurate in transition months in spring and fall. A suggested improvementis an updating recursive MOS system that is dependent on time of day.
展开▼
