Ultra‐short‐term wind speed forecasting method based on spatial and temporal correlation models

An ultra‐short‐term wind speed forecasting method based on spatial and temporal correlation models is proposed. Firstly, the temporal correlation model was studied. The generalised distance of wind speed time series is defined based on the dynamic time warping distance and correlation coefficient. Samples with similar wind dynamic patterns are classed into the same group. Time series in one group are depicted with a representative time series which is modelled by an autoregressive moving average (ARMA) model. For time series which cannot be classed to any of these groups, a special group was defined. It was modelled by artificial neural network (ANN). Secondly, the spatial correlation between the target wind farm and the reference wind farm was investigated with the meteorological data taking into consideration. The grey prediction method is utilised to predict the wind speed. Finally, the combined ultra‐short‐term wind speed forecasting method is proposed. It takes into account the inherent time correlation of the target wind speed and its spatial correlation with reference wind farm. The wind speed prediction of wind farms in northwest USA verifies the effectiveness and feasibility of the proposed method.