Wind power prediction method based on regime of switching kernel functions

The fluctuation of wind cause threat to power grid, this paper proposed a wind power prediction method to improving this situation. The proposed method is based on time series and regime of switching kernel functions. First, the mutual information method and the false nearest neighbor method were used to calculate parameters to reconstruct the original data. The recurrence figure and the Lyapunov exponent were applied to verify that the time series data was from a chaotic system. Then, this paper proposed a prediction method based on the kernel function and also a switching regime based on the support vectors machine. The new prediction method combining these two parts was proposed to predict wind power. The comparison of wind power prediction by the proposed method and traditional methods were present, the results validated that the proposed method is feasible to predict wind power, and that the precision of prediction is improved, which will be useful for the future analysis of wind power.