Estimation of confidence intervals for wind forecasts

In the Wind-Energy Market, the demand for confidence intervals of the supplied wind- and power-forecast is rapidly increasing. This quantification of the reliability of the forecast helps the vendor to estimate the probability of forecasted power. Especially low power production is of interest -- which may be caused either by too low wind speeds, or by shut-off of the park due to too high wind speeds.

MeteoGroup has developed a technique to accurately estimate the confidence intervals of the wind speed, the wind direction and the power. These intervals are dynamically based, i.e., they depend on the observed synoptic situation. The estimation consists of the following parts: • The optimal forecast is derived by combining multiple (global and regional) models. • All 51 EPS members of the ECMWF are used to capture the uncertainty in the forecast of the wind speed and -direction • Systematic biases between the ECMWF model and the observations are removed • The optimal forecast of the power, as well as its uncertainty range, are derived by nearest-neighbor resampling from the observed power curve.

This advanced method enables MeteoGroup to supply accurate forecasts as well as realistic uncertainty ranges, both for wind speed, wind direction and power output. The method can be applied to individual turbines as well as to wind parks.