Abstract: A method based on the gray theory and identification model was proposed to predict the short-term wind power generation. GM(2,1) model was applied for establishing a wind speed prediction model with an iterative step. After the wind speed prediction procedure, a finite impulse response moving average nonlinear Hammerstein model was used in the modeling between wind speed and wind power generation. By adopting the stochastic gradient searching theory, a wind power generation forecasting algorithm was proposed. The proposed simulation shows that the presented method can forecast the real time power generation of wind turbine and raise the accuracy of the wind power prediction, and the simulation that uses the actual data from real wind farm improves the practical applicability of proposed Grey-Identification model.

Key words: wind speed prediction, GM(2, 1) model, FIR-MA identification model, stochastic gradient algorithm