Ultra short-term combined forecasting of wind power with NAR neural network based on spatio-temporal correlation

Considering the spatial distribution correlation of the wind farm units, a combination prediction method based on the dynamic time series neural network(NAR) for the ultra-short-term power of wind farms is proposed. Firstly, the spatial correlation of output characteristics of wind turbine generators is analyzed by Empirical Orthogonal Function (EOF), and the wind turbines are classified according to the spatial characteristic contribution rate. Secondly, the NAR prediction model is established based on the mean value of original power time series from the wind units. Then the wind power prediction results are combined to get a total number of wind farm power forecasting value. A case study of a wind farm in northern China is carried out. The new model is compared with the single wind power series NAR prediction model and the typical ARMA(2, 2) prediction model. The comparison results verify the validity of the prediction results. © 2020, Solar Energy Periodical Office Co., Ltd. All right reserved.