Activities related to wind and wave energy forecasting at ECMWF

The successful integration of renewable energies into existing power systems and electricity markets partly relies on our ability to predict the output of renewable energy generators at different spatial and temporal scales. Over the last few years, the crucial role of numerical weather prediction in this field has been increasingly recognized. This is particular true for probabilistic weather predictions, generated either using ensemble prediction systems, statistically dressed single forecasts, or the combination of the two.

Since 1992, ECMWF has been at the front in the development of ensemble-based, weather prediction systems, and of applications of probabilistic forecasts in weather forecasting, hydrology, the energy sector and finance. In the past 2 years, in particular, ECMWF has been involved in an EU-funded project (SafeWind) that aimed to assess the value of ensemble-based probabilistic forecasts in wind energy applications. Besides this project, collaborations have been initiated related to the wave energy application.

As part of this project, an extensive verification of forecast products (both based on single high-resolution and ensemble forecasts) against observations has been performed, with forecast accuracy assessed taking into account observational uncertainty. This verification study is a first step towards the development of calibration/combination techniques that can lead to an optimal combination of high-resolution and ensemble probabilistic forecasts of near-surface variables. The final objective of this work is to combine both products in order to obtain an optimal probabilistic forecasting skill from 3-hour to 15-day ahead. During this talk, some recent results obtained within the SafeWind project will be discussed. In particular, the impact of the recent change in horizontal resolution on the skill of ensemble forecasts of wind speed will be presented. The improved skill of ensemble forecasts of near-surface winds over the European domain when recalibrated against observations will also be discussed.