基于混合聚类方法的电力系统负荷概率模型

A Probabilistic Model for Power System Load Based on Hybrid Clustering Method

针对利用k-均值聚类算法形成发电系统充裕度评估中的聚类负荷模型时存在的聚类中心初始值和聚类数难以确定的问题,提出根据负荷水平对充裕度指标的贡献度,将负荷曲线分成高贡献度、中等贡献度、低贡献度等分区,分别采用层次聚类、均值-标准差、随机法来选择各分区中聚类中心初始值;定义改进效率指标,将改进效率作为收敛条件确定聚类数.利用本文方法所得的聚类负荷模型,采用状态抽样法计算IEEE RTS79电力系统可靠性测试系统的发电系统充裕度指标.算例结果表明,同采用基于传统k-均值聚类方法的负荷模型结果相比,基于混合聚类方法得出的负荷模型的计算结果更精确,收敛速度更快.

K-means clustering algorithm is often employed to deduce the clustering load model corresponding to the annual chronological load, but it is difficult to determine the initial value of the clustering centers and clustering number. A novel load clustering method is presented in this paper. The load duration curve is divided into high-contribution, moderatecontribution and low-contribution sections according to the load level＇s contribution to the adequacy indices. And the initial values of clustering centers for each section are also obtained by using hierarchical clustering method, mean-standard method and random method. The paper.defines improved efficiency index, which is used as a convergence condition for determine clustering number. By employing state sampling method, the clustering load model generated from the proposed method is applied to the IEEE RTS79. Results of the case studies show that the clustering load model from the proposed method is more accurate and has a faster convergence rate than the load model which is based on conventional clustering methods.