Research on LS-SVM wind speed prediction method based on PSO

The impact on LS-SVM prediction model accuracy of the embedding dimension d and time delay τ were studied, which were used as a space reconstruction parameter. And LS-SVM prediction method based on PSO parameter optimization was proposed. In this method, particle swarm algorithm was used to optimize the embedding dimension d, time delay τ and other model parameters (regularization parameter γ, kernel function parameter σ), and then established prediction model. 2 groups of wind speed were predicted by using this method. The prediction error is about 5.79% and 7.33%, and the error of the contrast method (optimize γ, σ only) is 8.22% and 11.10%. The results show that the optimal selection of d, τ, γ, σ is necessary. Compare with the comparison model, this method can greatly improve the prediction accuracy. © 2016 Chin. Soc. for Elec. Eng.