Abnormal Detection of Electric Security Data based on Scenario Modeling

Abstract

Abnormal detection is an important task to ensure the stable running of the electric power system. In this paper, an abnormal detection framework for electric security data is proposed based on scenario modeling that supplies the descriptions of typical security issues. In this framework, the electric power system security-related data is first collected from different resources in real time. Then we create a scenario model based on the historical data for the typical security issues such as warning, system error or fatal error using the association rule mining algorithms. Then multiple scenario models are tagged with indicators that are observable from the sensors or monitors. Finally, the real-time data is analyzed to detect the defined indicators based on which the possible security scenarios are predicted and corresponding operation or protection will be activated automatically. The experiment results suggest that the proposed method can effectively identify and prevent the many of the common issues in the electric system.