In most operational wind power forecasting systems statistical methods are applied to map wind forecasts from numerical weather prediction (NWP) models into wind power forecasts. NWP models are complex mathematical models of the atmosphere that divide the earth’s surface into a grid. The spatial resolution of this grid determines how accurate meteorological processes can be modeled and thereby also limits forecast quality. In this study, two global and four regional operational NWP models with spatial horizontal resolutions ranging from 1 to 32 km were applied to make wind power forecasts up to 66 hours ahead for one offshore and two onshore Norwegian wind farms. A statistical meta-Gaussian method was applied to generate both probabilistic and deterministic wind power forecasts based on the NWP model wind forecasts. The experiments showed that the regional NWP models with higher resolution did not result in better wind power forecasts for these wind farms. In fact, the best wind power forecasts were obtained using one of the coarsest global NWP models.
|Number of pages||27|
|Publication status||Published - 2017|