Short-term wind speeds prediction of SVM based on simulated annealing algorithm with Gauss perturbation

Abstract

Abstract Wind speed prediction is an efficient means to reduce the downside effects of large-scale wind power generation on the grid, but the behaviour of wind speeds is nonlinear and non-stationary, which yields adverse challenge for its prediction. This work proposes a method of prediction for short-term wind speed, which makes Simulated Annealing Fruit fly Optimization Algorithm based on Gaussian Disturbance (GDSAFOA) to optimize the Support Vector Machine (SVM). In the method, Grey Relational Analysis (GRA) is used to select the factors which influence wind speeds prediction. A time series of wind speeds is decomposed by the Ensemble Empirical Mode Decomposition (EEMD). The wind speeds predication is the linear combination of the SVM and the dynamic neural network model based on the nonlinear autoregressive models with exogenous inputs (NARX). This method is applied for the model with wind speeds data measured from a wind farm in China’s Shanxi Province, where results exhibit that the proposed method is feasible and competitive.