Transient stability multi swing step-out prediction with online data mining

Recently, electric power systems become more complex and that makes it more difficult to control power systems. Due to development of WAMS(Wide Area Measurement System), power system data are available online. In this paper, we propose a novel method that can predict transient stability multi swing step-out using anomaly detection with data mining. Especially we focus our attention on the theory of ChangeFinder which uses SDAR algorithm and the two step learning model. The forgetting parameter r used in SDAR is set by supervised learning. Active powers obtained by transient stability simulations are inputted to ChangeFinder and the proposed method can detect multi swing step-out. We verify the validity of the proposed method by simulations on the IEEJ 10 machine 47 bus-system.