Intelligent Computing in Smart Grid and Electrical Vehicles

Wind power is influenced by multivariable, which usually shows complex nonlinear dynamics. Therefore the wind power is hardly described and traced by single variable prediction model; the precision of which decreases while it contains uncorrelated or redundant variables. The approach is proposed to reconstruct the phase space of multivariable time series and then predicate wind power. First, the delay time of single variable time series is selected by mutual information entropy, and then the embedding dimension of phase space is extended by the false nearest neighborhood method, which can eliminate the redundancy of reconstructed phase space from low space to high space. Then, the vector is utilized as input to predicate the wind power using the radial basis function neural network. Simulation of wind predication of Shanghai wind farms, show that the proposed method can describe the nonlinear system by less variables, and improve the precision and sensitivity of prediction.