基于小世界优化的支持向量机风电功率预测

WIND POWER FORECASTING BASED ON SMALL-WORLD OPTIMIZATION AND SUPPORT VECTOR MACHINE

采用基于实数编码的小世界优化算法(RSWOA)对SVM的惩罚因子C和核函数参数σ进行优化选取,使其具有收敛速度快和全局寻优的优点,提出基于实数编码小世界优化算法的支持向量机改进模型(RSWOASVM)。将该模型应用于实际风场的风电功率预测中,研究表明,RSWOA能快速准确找到SVM模型参数的全局最优解,进而可使RSWOA-SVM改进模型取得较理想的预测精度。

Generalizing ability and performance are largely dependent on the choice of key parameters of support vector machine （SVM）. The real-coding small-world optimization algorithm （RSWOA） was employed to optimize penalty factor C and kernel parameter σ of SVM in order to make it achieving fast convergence and global optimum. Thus, the improved SVM based on real-coding small-world optimization algorithm （RSWOA-SVM） was proposed. The research results which the new method was applied to the prediction of actual wind power showed that RSWOA can quickly find the global optimum of SVM parameters, and further make the RSWOA-SVM to achieve more perfect forecasting accuracy of wind power.