Analysis of wind power productions by means of an analog model

• Wind power output estimations are obtained by means of the analog model.
• Wind power data involve both individual wind farms and wind farm clusters.
• Large-scale database dimensionality is reduced by using multivariate methodology.
• Results are shown in terms of deterministic and probabilistic verification statistics.

The purpose of this work is to evaluate the performance of an analog model on day-ahead forecasting of wind power production over large European regions based in Ireland, Denmark and Germany. To do this, several data sets have been used: sea level pressure field over the North Atlantic and wind power outputs from individual wind farms and from wind farm clusters. The analog method uses Principal Component Analysis to reduce the dimensionality of the large-scale atmospheric database. Then, the analog method is based on the finding in the historic sea level pressure database, a principal component subset of large-scale atmospheric patterns that are the most similar to a large-scale atmospheric pattern used as input. Similar atmospheric situations to a particular atmospheric situation to be modeled have been determined and from them, different wind power outputs have been estimated. Several deterministic and probabilistic results are shown. Results of bias, spatial correlations and root mean squared errors between the estimated and observational wind power outputs are displayed. Concerning wind farm data set, the analog method improves both climatology and persistence in the Danish test case. The probabilistic results are shown by means of Brier Skill Scores and reliability diagrams. Danish test case shows pretty good BSS results with underestimation of the observational wind power frequencies in the reliability diagrams. For aggregated data sets, the model performing improves climatology in both Danish and German test cases, showing the latter better results than the former. A comparison between the two Danish databases, wind farm and aggregated data, gives as result higher BSSs for aggregated data than for the wind farm data set in high wind power outputs. The process used in this work to estimate wind power productions based on finding analogs in a previously reduced large-scale atmospheric data has proven to be a good technique to analyze the effect of the regional wind climate contribution to the daily wind output prediction.