Performance and sensitivity of a rapid update numerical weather prediction system for 0-6 hr ahead wind power production forecasting

A customized forecast system based on the Advanced Regional Prediction System (ARPS) NWP model operated in rapid update cycle mode and a regime-based model output statistics (MOS) scheme has been developed for the purpose of improving 0-6 hr ahead prediction of wind power production. The primary objective of the system is the prediction of large changes in wind speed over short time periods that cause significant changes in power production, which are known as wind ramps. The ability to reliably forecast these events will facilitate the addition of larger amounts of wind power onto the grid.

The system is being applied for wind farms in several regions of the central US. ARPS simulations with a 1-hour or 2-hour update frequency and horizontal resolutions of 6 to 10 km are used to produce deterministic high frequency forecasts of the flow within the layer from 50 m to 100 m above ground level. The system employs the ARPS three-dimensional variational (3DVAR) scheme to assimilate data from a variety of sources including surface mesonet data, wind profilers, Doppler radars and meteorological observations from the wind farms themselves. As part of the development and evaluation process for this forecast system, an investigation of the impact of grid resolution, type of assimilated data and the formulation of the model boundary layer parameterization on performance of the ARPS low-level wind forecasts (before and after the application of MOS) was examined by executing parallel runs with different configurations. An overview and analysis of forecast performance comparisons from the parallel runs will be presented. The analysis of the results provide considerable information into the relative importance of these factors for short-term 50-m to 100-m wind forecast performance as well as insight into the situations in which forecast uncertainty is the largest. The initial results indicate that vertical processes and convection are associated with the largest short-term forecast uncertainty.