In the paper, the chaos least squares support vector machine algorithm (Chaos-LS-SVM) is applied. To conduct uncertainty analysis of wind power forecasting, two forecasting algorithms of the probabilistic uncertainty analysis based on the Monte Carlo method and the quantile regression analysis based on Chaos-LS-SVM are discussed. The effectiveness and superiority of the two uncertainty analysis methods in the confidence level of 95%, 90%, and 85% are discussed by simulation analysis, and the confidence interval is given in the corresponding confidence level. The prediction interval coverage probability (PICP) and the prediction interval normalized average width (PINAW) of the two uncertainty methods are compared. In the time scale of 1h-ahead, 4h-ahead, and 6h-ahead, the probabilistic uncertainty analysis based on the Monte Carlo method is suitable. In the time scale of 24h-ahead, the quantile regression analysis based on Chaos-LS-SVM is superior.

本文应用了混沌最小二乘支持向量机算法（Chaos-LS-SVM）。针对风电功率预测的不确定性分析，讨论了基于蒙特卡罗方法的概率不确定性分析和基于Chaos-LS-SVM的分位数回归分析两种预测算法。通过仿真分析讨论了两种不确定性分析方法在95%、90%和85%置信水平下的有效性和优越性，并在相应的置信水平下给出了置信区间。比较了两种不确定性方法的预测区间覆盖概率（PICP）和预测区间归一化平均宽度（PINAW）。在提前 1 小时、提前 4 小时和提前 6 小时的时间尺度上，基于蒙特卡罗方法的概率不确定性分析是合适的。在提前 24 小时的时间尺度上，基于 Chaos-LS-SVM 的分位数回归分析更胜一筹。