This paper describes a refinement of wind speed prediction methods in order to enhance their accuracy for wind energy applications. Specifically, techniques used to downscale raw forecasts from numerical weather prediction models are investigated. Many downscaling techniques have been proposed, however most of these rely on wind speed data while ignoring a potentially valuable source of information, namely wind direction. In this paper, we incorporate
wind speed and direction into three downscaling methods: linear model output statistics; feedforward artificial neural network (ANN); and Kalman filter (KF). We apply the techniques to downscale outputs of a global numerical weather prediction model to six test locations in Ireland for which wind speed and direction measurements were available.
While classical downscaling methods require large sets of historical data in order to be trained, the KF has the potential to rapidly estimate the bias that needs to be added to the raw forecasts in order to provide the best fit possible to local observations. Comparing the results of the three downscaling methods, it is shown that while the levels of prediction accuracy attainable with the KF are similar to classical techniques, the amount of data required to parameterise the KF is much less than for other techniques. The KF has a further advantage over the ANN in that it does not require offline parameterisation. However, in this study, the ANN performance was more satisfactory in reducing prediction errors.