Wind speed and wind energy forecast through Kalman filtering of Numerical Weather Prediction model output

Abstract

Despite the major progress made by Numerical Weather Prediction (NWP) in the last decades, meteorological models are usually unable to provide reliable surface wind speed forecasts, especially in complex topography regions, because of shortcomings in horizontal resolution, physical parameterisations, initial and boundary conditions. In order to reduce these drawbacks, one of the most successful approaches is the Kalman filtering technique, which combines recursively observations and model forecasts to minimise the corresponding biases. In meteorology, Kalman filters are widely used to improve the prediction of variables characterised by well-defined cyclicities, whereas the evolution of wind speed is usually too irregular. In the present paper, the Kalman filter is analysed in order to find the best configuration for wind speed and wind power forecast. The procedure has been tested, in a hindcast mode, with 2-year-long data sets of wind speed provided by a NWP model and two anemometric stations located in the eastern Liguria (Italy). It is shown that, tuning time step and forecast horizon of the filter, this methodology is capable to provide significant forecast improvement with respect to the wind speed model direct output, especially when used for very short-term forecast. In this configuration, Kalman-filtered wind speed data have been used to forecast the wind energy output of the nearby wind farm of Varese Ligure. After 2years of testing, the percentage error between simulated and measured wind energy values was still very low and showed a stable evolution.