Relative Benefits of Various Meteorological Forecasting Inputs for Wind Power Prediction

Accurate wind power forecasts are essential for the optimal integration of wind energy into power systems. Usually, forecasts for lead times beyond approximately 6 hours are based on numerical weather predictions that use physical considerations to predict future states of the atmosphere. A variety of different numerical weather prediction models exists that differ, e.g., in spatial or temporal resolution or extent. Furthermore, many weather centers provide ensemble forecasts to estimate the forecast uncertainty or compute large reforecast data sets of forecasts for past dates.
This study compares different numerical weather prediction systems for wind power forecasting. Statistical post-processing is applied to eliminate systematic model errors.
The results show partly substantial quality differences between different numerical models where a higher spatial resolution not necessarily implies a better performance. Furthermore, the long reforecast datasets can have clear advantages for training statistical post-processing models.