Dealing with Power Curve Uncertainty — Evaluation of Various Wind (power) Calibration Techniques

The conversion from wind to turbine generated power is widely done via so-called ‘power curves'. The power curve, supplied by the turbine manufacturer, gives the static relationship between mean wind speed and power, allowing a straightforward computation of power from wind forecast. However, it is commonly known that wind turbine performance declines with age due to mechanical wear and tear. This loss of efficiency is not represented in the power curves and may therefore become a source for systematic forecast errors with progressing turbine age. Furthermore, site characteristics like the typical air density and the position of the turbine within the farm can influence the power output. This may also lead to forecast errors when the manufacturer's power curve is used for the translation of Numerical Weather Prediction (NWP) model wind to turbine generated power.

MeteoGroup has applied statistical techniques based on Weibull calibration and Kalman filters to address these systematic errors in the site-specific wind power forecast. Both Weibull calibration and Kalman filtering are, on average, successful in removing the systematic errors (bias) and thereby lowering the MAE. However, the verification scores vary greatly among the wind farms. It is therefore recommended that a highly flexible power bias correction solution should be put in place.