Out-of-step detection based on pattern recognition

Abstract This paper presents a new approach for out-of-step detection of synchronous generators with the ability to initiate early tripping for non-recoverable swings while avoiding tripping on recoverable swings. The approach is based on the K -means clustering pattern recognition technique. The capabilities of the developed algorithm are tested through computer simulation for a typical case study. While being feasible for on-line implementation, the introduced algorithm shows high classification performance.