基于混合聚类分析的电力系统网损评估方法

面向具有海量性、多元数据类型特征的电网运行方式数据集,如何快速、准确地实现电力系统网损评估将是一个重要研究点。采用数据挖掘和典型场景模拟思想,提出了一种新颖的基于混合聚类分析的网损评估方法。该方法首先面向网损评估确定聚类属性;其次根据各聚类属性的数值类型异同将原聚类问题分解为两个聚类子问题,进而在充分考虑电力数据特点的基础上,分别选取划分聚类算法和层次聚类算法对其进行聚类分析,并集成各子问题的聚类结果;最后基于混合聚类结果生成电网的典型运行方式集,以用于网损评估。以某省级电网为算例验证所提方法在网损评估中的有效性,结果表明基于所提方法的网损评估精度较高,计算效率较好,在工程实际中具有较强的实用性。

A Power System Network Loss Evaluation Method Based on Hybrid Clustering Analysis

Faced with the power system operation data set which has the characteristics of massive amounts of data and multiple numerical types, how to achieve a rapid and accurate network loss evaluation in power system will become an important issue. In accord with the data mining and typical scenario simulation method, a novel network loss evaluation method based on hybrid clustering analysis is proposed. The clustering attributes associated with network loss are determined from the power system operation data set first. Based on the division of different numerical types of clustering attributes, the original clustering problem is divided into two clustering sub-problems. In consideration of the characteristic of power data, the partitioning clustering method and hierarchical clustering method are adopted respectively to deal with the two clustering sub-problems. The clustering ensemble technique is further employed to achieve a mixed clustering result, which is finally utilized to generate a typical power system operation data set for the evaluation of network loss. The numerical result shows the proposed network loss evaluation method has good evaluation precision and computational efficiency, which can be effectively applied to engineering practice. This work is supported by National Natural Science Foundation of China (No. 51277141).