基于光滑自助法的风功率预测误差核密度建模方法

在风功率预测误差建模应用中,无偏交叉验证(UCV)和经验法则(ROT)是两种常用的非参数方法。然而,由于风功率预测误差中存在的尖峰厚尾,以及局部小样本特征,直接使用这两种方法会产生较大的泛化误差。为了使UCV和ROT在应用中发挥更好的作用,文章提出了一种基于光滑自助法的核密度估计方法。该方法利用了光滑自助法在分位数推断上的优势,通过修改平均积分平方误差(MISE)指标函数,实现了对基本估计方法的校正。该方法本质上是一种装袋方法,可以与任何基本的核密度方法结合使用。在实例仿真中,得到了SBUCV方法和SBROT方法的运行结果,并与UCV和ROT方法的结果进行了对比。仿真结果表明了该方法的有效性。

Kernel density estimation of wind power forecast error based on smooth bootstrap method

UCV(unbiased cross validation) and ROT(rule of thumb) are both common non-parametric methods to predict the error model of wind power system applications. However, these methods will produce a large generalization error because of the leptokurtic features and local small sample characteristics. In order to solve the above problems, a kernel density estimation method based on smooth bootstrap is proposed in this paper. This method is essentially a type of bagging method that could be used in combination with all kinds of kernel density estimation methods. It realizes the correction of the original estimation by taking advantage of the smooth bootstrap method in quantile inference and modifying the MISE(mean integrated squared error)indicator function. The simulation compares the running data of the comprehensive SBUCV(smoothbootstrap unbiased cross-validation)and SBROT(smoothbootstrap rule of thumb)methods with the results of the basic UCV and ROT methods. Results demonstrate the correctness and validity of the proposed method.