Short-term wind power prediction based on kernel function switching

作者：Ouyang Tinghui; Zha Xiaoming; Qin Liang^*; Xiong Yi; Xia Tian; Huang Heming

来源：Electric Power Automation Equipment, 2016, 36(9): 80-86.

DOI：10.16081/j.issn.1006-6047.2016.09.012

• Chaotic time series
• Chaotic time series prediction
• False nearest neighbor
• Kernel function
• Large-scale wind power integrations
• Mutual information method
• Short-Term wind power predictions
• Wind power predictions

A method of chaotic time series prediction based on the switching regime of kernel functions is proposed to further improve the performance of short-term wind power prediction for reducing the potential risk of power grid caused by the large-scale wind power integration. The mutual information method and false nearest neighbor method are applied to reconstruct the phase-space of original wind power series. The recurrence plot and the maximum Lyapunov value are used to verify that, the wind power series are from a chaotic system with certainty and randomness and the chaotic prediction is applicable. The implementation of chaotic time series prediction based on kernel functions is given and the training sample analysis proves it is better than the traditional prediction method. According to the training results, the support vector machine is proposed to train the switching regime of optimal kernel functions for future improving the prediction accuracy. As an example, the comparison among the error indexes based on the data from BPA website proves that, the prediction based on the kernel functions with switching regime can effectively realize the short-term wind power prediction with better performances.

• 出版日期2016