The invention discloses a quick robust classification method of preconceived faults for power system transient stability assessment and belongs to the technical field of power systems and automation thereof. The method is based on an extended equal area criterion and divides examples in a preconceived fault universal set into five classifications, i.e., stable, doubtfully stable, critical, doubtfully instable and instable by deeply mining different allowance information and comparison results thereof contained by transient stability analysis algorithms with different integration steps, reflecting the time-varying degree of research examples and forming identification rules for different classifications of transient stability severity in combination with fault information. The quick robust classification method of preconceived faults for power system transient stability assessment can realize high-efficiency, reliable and reasonable classification of different classifications of transient stability severity of examples in the preconceived fault universal set, and has a great theoretical and engineering significance to deeply coordinate online transient stability analysis accuracy and speed and solve the problem of transient stability analysis taking account of uncertainty factors.