Probabilistic Forecasts of Wind Speed: Ensemble Model Output Statistics by using Heteroscedastic Censored Regression

Abstract

As wind energy penetration continues to grow, there is a critical need for probabilistic forecasts of wind resources. In addition, there are many other societally relevant uses for forecasts of wind speed, ranging from aviation to ship routing and recreational boating. Over the past two decades, ensembles of dynamical weather prediction models have been developed, in which multiple estimates of the current state of the atmosphere are used to generate a collection of deterministic forecasts. However, even state of the art ensemble systems are uncalibrated and biased. Here we propose a novel way of statistically post-processing dynamical ensembles for wind speed by using heteroscedastic censored (tobit) regression, where location and spread derive from the ensemble. The resulting ensemble model output statistics method is applied to 48-h-ahead forecasts of maximum wind speed over the North American Pacific Northwest by using the University of Washington mesoscale ensemble. The statistically post-processed density forecasts turn out to be calibrated and sharp, and result in a substantial improvement over the unprocessed ensemble or climatological reference forecasts. Copyright (c) 2009 Royal Statistical Society.

A General Probabilistic Forecasting Framework for Offshore Wind Power Fluctuations

“…For instance, the additional improvement in the CRPS criterion for a simple AR model is about 7%-8%. These results are in line with those reported in [21][22][23] which showed the potential of using a truncated Normal distribution for wind speed and wind power prediction applications. Similarly, the use of the Generalized Logit-Normal distribution for Markov-Switching will be investigated with a particular focus on multi-step ahead forecasts.…”

“…In view of these limitations, truncated and censored normal distributions stand as appealing alternatives to the more classical Normal distribution. Recent developments that use the two former distributions applied to wind data include [22,23]. However, Markov-Switching models imply the computation of distribution mixtures.…”

Estimation of the Continuous Ranked Probability Score with Limited Information and Applications to Ensemble Weather Forecasts

“…with Grimit et al (2006) p i=1 ω i = 1, ω i=1,..., p > 0 Generalized extreme value: Y ∼ G E V (μ, σ, ξ ) Friederichs and Thorarinsdottir (2012) Generalized Pareto: Y ∼ G P D(μ, σ, ξ ) Friederichs and Thorarinsdottir (2012) Log-normal: ln(Y ) ∼ N (μ, σ ) Baran and Lerch (2015) Normal: Y ∼ N (μ, σ ) Gneiting et al (2005) Square-root truncated normal: √ Y ∼ N 0 (μ, σ ) Hemri et al (2014) Truncated normal: Y ∼ N 0 (μ, σ ) Thorarinsdottir and Gneiting (2010) The reference of the original article where to find the formula is also given. Taillardat et al (2016) gathers the closed form expression of the CRPS for these and other distributions in Appendix A, this score for an ensemble is minimized if all the members x i equal the median of F, which is obviously not the purpose of an ensemble.…”

On the skill of various ensemble spread estimators for probabilistic short range wind forecasting

“…The EPS dyn is further calibrated by applying a cut-off variant of the nonhomogeneous Gaussian regression (NGR) technique (Thorarinsdottir and Gneiting, 2010;Gneiting et al, 2005;Hagedorn et al, 2008;Kann et al, 2009). This technique statistically calibrates the mean and the ensemble variance by minimizing the Continuous Ranked Probability Score (CRPS) within a certain training period.…”