一种短期光伏出力的区间预测方法

摘要

In allusion to the problem that deterministic point prediction can't meet the requirement for dispatching of the large-scale PV grid-connection in precision,this paper presents a kind of prediction method for PV output intervals.Consid-ering strong fluctuation characteristics of original data of PV power,it adopts the variational modal decomposition (VMD) method to decompose the original data into several sub-sequences,and according to sample entropy theory,it reorganizes the sub-sequences with higher complexity into the fluctuating component.Then by means of Gaussian process regression(GPR), it makes a prediction on this component so as to obtain its fluctuation interval.In consideration of inherent defects of GPR, the paper uses crisscross optimization (CSO)algorithm to improve the optimization process of super parameters.It regards the other sub-sequences with lower complexity as stable components of PV output and directly uses support vector machine (SVM)method for deterministic prediction.Finally,by reconstructing predictive values of various components,it obtains prediction results of PV output intervals.%确定性的点预测在精度上无法满足大规模光伏并网的调度需求,基于此,提出一种光伏出力区间预测方法.针对光伏功率原始数据的强波动特性,采用变分模态分解(variational model decomposition,VMD)方法将其分解为若干个子序列,并依据样本熵理论,将复杂度较高的子序列重组为波动分量 S,采用高斯过程回归(Gaussian process regression,GPR)法对分量S进行预测,得到其波动区间.考虑到 GPR 本身固有的缺陷,采用纵横交叉(crisscross optimization,CSO)算法对它的超参数寻优过程进行改进,而复杂度相对较低的其他VMD 子序列代表光伏出力稳定分量,因此,采用支持向量机(support vector machine,SVM)法直接对它们进行确定性预测,最后通过重组各分量的预测值,得出光伏出力的区间预测结果.