The role of regimes in short‐term wind speed forecasting at multiple wind farms

Abstract

Large‐scale integration of wind energy into electric utility systems requires accurate short‐term wind speed forecasts. At these horizons, statistical models that account for spatial and temporal information have demonstrated improved accuracy over both physical models and statistical models that ignore spatial information. Off‐site information can be incorporated by modelling wind speeds conditional on a set of regimes that capture the predominant wind patterns within a geographic region. Identifying these regimes is a crucial model‐building step. Herein, we propose a new forecasting method that relies on regimes identified by fitting a Gaussian mixture model (GMM) to the wind vector, and we build regimes based on a single site, a local average of sites, and a region‐wide average. We compare the performance of the models with GMM‐identified regimes with three state‐of‐the‐art reference models that each account for wind regimes differently. The models are evaluated at 30‐minute, 1‐hour, and 2‐hour ahead horizons at ten sites across the Pacific Northwest. GMM regimes based on local information produce the best forecasts and have a significantly improved accuracy at a region‐wide level over the state‐of‐the‐art models. Even greater improvements are achieved when an average of the forecasts produced by each method is constructed. Copyright © 2015 John Wiley & Sons, Ltd.

Improved very short‐term spatio‐temporal wind forecasting using atmospheric regimes

“…Considering mixtures of regression models rather than discrete switching may also yield improvement. ()…”

“…Large‐scale pressure differences that result in either easterly or westerly winds being channelled through the Columbia Gorge define the regimes, and regimes are switched between based on recent observations of wind direction for forecasting. A more general approach is developed in Hering et al and Kazor and Hering by fitting a mixture of bivariate normal distributions to the wind vector, which may be a spatial average, and defining regimes by the mixture components. The identification of regimes based on multiple measurement locations has been studied in Burlando et al and Kazor and Hering .…”

A Spliced Gamma-Generalized Pareto Model for Short-Term Extreme Wind Speed Probabilistic Forecasting

“…An important step to fit our off-site latent model described in Section 2.3.1 is to select a suitable set of off-site predictors N s , s ∈ S. Here, we develop a data-driven approach for identifying dominant wind directions, which we then use subsequently to automatically choose the off-site predictors. This procedure is similar in spirit to that in Kazor and Hering (2015) that their distance to s is less than some maximum distance d max . We selected the number of components for the mixture of von Mises distributions via the Bayesian Information Criterion (BIC), also guided by the wind roses displayed in Figure 1 of the Supplementary Material.…”

“…Models that incorporate both temporal and spatial correlations in the form of off-site information (i.e., information that is not collected at the site of interest, but at neighboring sites) have been found to increase accuracy over conventional time series models. In this framework, Alexiadis et al (1999) use off-site predictors to improve wind speed and wind power forecasts, while Gneiting et al (2006), Hering and Genton (2010), and Kazor and Hering (2015) assume that wind speeds follow a truncated normal distribution with regime-dependent mean and variance. See Zhu and Genton (2012) for a review on statistical wind speed forecasting models.…”

Bridging asymptotic independence and dependence in spatial extremes using Gaussian scale mixtures

“…Selected data comprise up to 6504 hourly observations at each site, with about 8% of values missing. Owing to important East-West pressure gradients in this region and the special nature of the orography, wind patterns are mainly characterized by easterly and westerly winds(Kazor and Hering, 2015). A wind rose for the data at the 12 stations reveals that extreme winds blow mostly from the West or South-West, suggesting that simple anisotropic models might perform well; seeFigure 6.More details about data and monitoring stations are in the Supplementary Material.Hourly wind speeds show strong temporal dependence.…”