Back to the programme printer.gif Print

Delegates are invited to meet and discuss with the poster presenters in this topic directly after the session 'Forecasting: Maximizing grid deliverability and leading your business processes to profitability' taking place on Thursday, 13 March 2014 at 11:15-12:45. The meet-the-authors will take place in the poster area.

Luca Delle Monache National Center for Atmospheric Research (NCAR), United States of America
Luca Delle Monache (1) F P
(1) National Center for Atmospheric Research (NCAR), Boulder, United States of America (2) Vattenfall, Fredericia, Denmark

Printer friendly version: printer.gif Print


On the value of uncertainty quantification and probabilistic wind power predictions


Among the limiting factors to the penetration of wind energy is the variable nature and limit to predictability of wind speed. We will introduce a novel technique to generate short-term wind power predictions, the analog ensemble (AnEn, Delle Monache et al. 2013; Alessandrini et al. 2013) and estimate the economic value of AnEn-based short-term probabilistic predictions and uncertainty quantification for real-world examples over Europe and USA.


In AnEn, for each forecast lead time and wind farm, the ensemble wind power prediction is constituted by a set of past measurements of power. These measurements are those concurrent to past deterministic Numerical Weather Predictions (NWP) for the same lead time and wind farm, chosen based on their similarity to the current NWP forecast. The economic impact of AnEn probabilistic forecast and uncertainty quantification for real-world examples is evaluated introducing a model of an electricity market, which simulate a system where penalties must be paid by the producers of the energy on unbalancing.

Main body of abstract

Wind power predictions can be categorized in two main groups: deterministic and probabilistic. A deterministic forecast consists in a single value for each time in the future for the variable to be predicted. Probabilistic forecasting provides probability density functions (PDF) from which probabilities of future outcomes of events can be estimated. This implies also providing information about uncertainty in addition to the commonly provided single-valued power prediction.

A deterministic prediction can provide useful information for decision-making. Its utility, however, is fundamentally limited as it represents only a single plausible future state of the atmosphere from a continuum of possible states, which result from imperfect initial conditions and model deficiencies that lead to nonlinear error growth during model integration (Lorenz 1963). Accurate knowledge of that continuum, the forecast PDF, provides considerably more utility to decision-making (e.g., Hirschberg et al. 2011).

We will present real-world examples of probabilistic wind power predictions and uncertainty quantification produced with the AnEn technique at a wind farm in Sicily, Italy, over a 505-day period, and wind farms located in Central USA over a period of several months. We quantify the economic value of AnEn predictions with a model of an electricity market and compare it to the value of deterministic power forecasts. We demonstrate that the optimal day-ahead market bid for a wind energy producer (Roulston et al. 2002; Zugno et al. 2013) is a quantile of the wind PDF that is different from the deterministic prediction.


Park (2011) reported that a 20% reduction of mean-absolute-percentage-error resulted in $2.5 million savings in one year for wind predictions over a facility with a 1561 MW capacity. These deterministic predictions were based on a power forecasting system developed at U.S. National Center for Atmospheric Research, which is also been used in the experiments presented here as one of the deterministic predictions used to generate AnEn. In our experiments with real-word data from wind farms in Europe and USA, we found that the economic benefit of a probabilistic approach compared to a deterministic one corresponds to a 10-20% increase of the annual income.

Learning objectives
- Introduction to a novel technique, the analog ensemble, for the generation of short-term probabilistic power predictions and uncertainty quantification;
- Real-world examples of the economic impact of this technique;
- Demonstration on how this impact can be estimated with an electricity market model.