Real time transient instability assessment based-on Bayesian theory

This paper presents a real time method for Out-of-Step (OS) detection of synchronous generators of Wide Area Network (WAN) based on Bayesian theory. For distinguishing between power swing and out-of-step conditions a series of event base features is introduced and used for Bayesian training testing. For sample generation a series of measurements are taken with 50ms sampling windows under various faults including operational and topological disturbances. The proposed method is used on New-England 345 KV test system and the simulation results are used as input-output pairs data for Bayesian learning and constructing. The merit of out-of-step proposed protection scheme lies in adaptively and robustness of features.