On the quality and value of probabilistic forecasts of wind generation

Abstract

While most of the current forecasting methods provide single estimates of future wind generation, some methods now allow one to have probabilistic predictions of wind power. They are often given in the form of prediction intervals or quantile forecasts. Such forecasts, since they include the uncertainty information, can be seen as optimal for the management or trading of wind generation. This paper explores the differences and relations between the quality (i.e. statistical performance) and the operational value of these forecasts. An application is presented on the use of probabilistic predictions for bidding in a European electricity market. The benefits of a probabilistic view of wind power forecasting are clearly demonstrated. © Copyright KTH 2006.