Spatio‐temporal propagation of wind power prediction errors

Abstract

International audienceThe increasing concentration of wind farms in some parts of the world calls for a new descriptive framework of power fluctuation that can summarize spatio-temporal characteristics of the wind power production process. In the mean time, this high number of measurement devices has great potential for informing or alerting about upcoming front or phase errors. In this paper, we shed light on the spatio-temporal characteristics of wind power forecast errors. We justify and use two strategies for obtaining the main direction and the speed of propagation. The first of these relies on an analysis of the local maxima of the correlation structure, and the second builds upon a planar wave modeling. The connection is made with meteorological parameters. Our analysis is presented so as to be easily reproducible in other cases

Probabilistic Forecasts of Wind Power Generation Accounting for Geographically Dispersed Information

“…This difference with the results in Ref. [28] could be explained by a higher spatial persistence and homogeneity of wind field dynamics over waters than over lands, where the terrain roughness is known to be a very influential factor.…”

“…In other words, if the reference sites are rather remote and it takes longer than k for the information to propagate from the reference to the target point, then a snapshot of the past errors (t − h) should be used as explanatory variables. If on average it takes less than k for the information to propagate, one should use the corresponding snapshot taken at time t. Preliminary data analysis (for example, cross-correlation analysis of the forecast errors) can be used to get a hint on the average speed of error propagation over the territory [28]. Further in this work we focus on the case with h = 0, i.e we use the latest available information as explanatory variables.…”

“…[28] it was found that correlations in wind power forecast errors between two wind farms were very small for distances larger than 50 km, over the Western Denmark area. Our findings are consistent with these results since most wind farms that contribute to improve wind power predictability at Nysted are located within such range (see Fig.…”

“…Owing to their geographical dispersion, wind farms comprise a dense network of atmospheric sensors capable of capturing valuable information on the spatio-temporal propagation of meteorological systems, and thereby on the propagation of wind power forecast errors [27], [28]. Historically, most stateof-the art wind power prediction systems are optimized and run locally, for a single region, site or wind farm of interest, using on-site information only (e.g., meteorological forecasts, historical measurements of wind power) [24].…”

Space-Time Trajectories of Wind Power Generation: Parametrized Precision Matrices Under a Gaussian Copula Approach

“…This difference in the dependency patterns can be partly explained by the fact that in Denmark the prevailing winds are westerly. Thus, forecast errors most often propagate from West to East, as discussed in, e.g., [15]. This means that usually zones A and N are influenced by the upcoming weather front simultaneously, while zone W is exposed to it earlier.…”