期刊名称：International Journal of Intelligent Systems and Applications
印刷版ISSN：2074-904X
电子版ISSN：2074-9058
出版年度：2019
卷号：11
期号：2
页码：25-35
DOI：10.5815/ijisa.2019.02.04
出版社：MECS Publisher
摘要：Power system contingency studies play a pivotal role in maintaining the security and integrity of modern power system operation. However, the number of possible contingencies is enormous and mostly vague. Therefore, in this paper, two well-known clustering techniques namely K-Means (KM) and Fuzzy C-Means (FCM) are used for contingency screening and ranking. The performance of both algorithms is comparatively investigated using IEEE 118-bus test system. Considering various loading conditions and multiple outages, the IEEE 118-bus contingencies have been generated using fast-decoupled power flow (FDPF). Silhouette analysis and fuzzy partition coefficient techniques have been profitably exploited to offer an insight view of the number of centroids. Moreover, the principal component analysis (PCA) has been used to extract the dominant features and ensure the consistency of passed data with artificial intelligence algorithms’ requirements. Although analysis of comparison results showed excellent compatibility between the two clustering algorithms, the FCM model was found more suitable for power system static security investigation.
关键词：Artificial intelligence;cascading outages;contingency screening and ranking;fuzzy c-means;k-means
