【摘要】 针对支持向量机方法在短系列电力负荷预测中存在空间划分参数的选择受主观因素影响的缺点,提出了谱分析和最小二乘支持向量机相结合的负荷预测方法。该方法采用谱分析预测实际发生最大电力负荷值的周期,根据周期来确定SVM的训练模型。该方法可有效地避免参数选择中的人为因素,提高预测精度。从实际算例可看出,除最后一个点位相对误差为8.67%外,其余点位的相对误差均低于±5%,实测值与预测值的拟合度较好,预测精度较高。更多还原

【Abstract】 For the shortcoming of the SVM using in short series of power load forecasting,namely the choice of space divided parameters is influenced by subjective factors,a novel method of power load forecasting combining the spectrum analysis with LS-SVM method is presented. This method adopts the spectrum analysis to forcast the cycle of actual maximum power load and determines the training mode of SVM based on the cycle.The subjective factors in preferences are effectively avoided,and the prediction accuracy is improved.The practical example shows that except the last point in which the relative error is 8.67%,the relative errors of other points are less than ± 5%.Our method provides a better fit to the predicted data and shows high accuracy.更多还原