The Influence of Wind Farms Aggregation Methods on the Efficiency of Short-Term Power Generation Forecasting

The stochastic nature of wind energy entails the need of accurately forecasting the output produced by wind farms for managing the wholesale electricity market. The power output produced by individual wind turbines united into wind farms may differ significantly as a consequence of wind turbines shadowing each other, effects of obstacles, etc. Instead of constructing a single model for the entire wind farm, several forecast models for wind turbine groups and individual wind turbines can be used to take the shadowing into account and improve the forecasting accuracy. The article presents a study of the ways in which individual wind turbines can be aggregated for achieving more efficient short-term forecasting of the wind farm outputs. A new wind turbines grouping technique is proposed, central to which is the hierarchical agglomerative clustering method based on dynamic time warping (DTW) distance measure. Owing to the use of the new technique, the forecast r.m.s error has been decreased by 8.7 % in comparison with the forecast made at the wind farm level.