Intermittent wind power production generates risk for utilities trading in short-term markets. Changes in wind power predictions cause price feedbacks on intraday and balancing markets. Extreme weather events can restrict wind power operations, e.g. high-wind shutdown, icing or curtailment. In order to manage and reduce theses complex weather risks, a holistic perspective on forecasting and process optimization is needed that considers the whole value chain from weather model output to final trading and dispatch decision.
We study an operational wind power forecasting system and demonstrate the importance of key elements for different example events. The wind power forecasting system consists of a multi model approach allowing for human intervention and bid optimization depending on latest market and weather developments. Wind power models are forced with weather data from multiple global weather models. Machine learning algorithms are used to mix forecast models depending on latest performance and prevalent weather regimes. Portfolio uncertainty is assessed utilizing an analog ensemble, that allows to derive trading strategies for day-ahead bidding. Ultra-short-term corrections are executed by a nowcasting algorithm using live data and artificial intelligence. We analyse the ability of the operational forecast system to deliver good intraday trading results and optimal dispatch decisions to mitigate imbalance costs and grid imbalance.
While scientific forecasting developments have traditionally a focus on improving forecast error statistics, an operational forecasting system requires a broader scope. For the decision in day-ahead markets it can be of high value to have a good quantification of uncertainty in the forecast (i.e. asymmetric risks). In intraday for instance, forecast delivery speed and correct prediction of the direction of change are often more important than the absolute error itself. Receiving observational data from wind farms in near real-time and having a robust and effective data validation can be decisive for managing short-term risks. Additionally, dedicated forecasts on curtailments, icing etc. are an essential element to manage financial risk adequately.
Requirements and challenges in wind power forecasting depend on the individual stakeholder. While grid providers rely on forecast accuracy, trading utilities focus on optimizing revenues in a highly weather dependent market environment and have therefore different requirements on the operational forecasting system.