Transient stability assessment by pattern recognition in the frequency domain

A pattern recognition methodology is presented which utilizes spectral monitoring of electromechanical oscillations to assess the transient stability of interconnected power systems. The proposed frequency-domain approach permits definitive recognition of unstable dynamic modes. This is used to design an adaptive linear classifier which, in the developmental implementation, virtually eliminates both false dismissals and false alarms. The result is an accurate and reliable decision-making system which performs transient stability monitoring and assessment in real time.