Identification of post-fault topology is critical to not only cascaded outage prevention but also prediction of post-fault dynamic behavior. Existing methods for identifying the tripped lines rely on steady state SCADA/PMU measurements, and therefore fail to accomplish their task within fractions of a second. This limits their application to online transient stability assessment. This paper presents a method for online identification of post-fault topology on the basis of information of the pre-fault operating conditions and synchronized measurements at all generator buses. The proposed technique identifies the tripped line with negligible delay, before the system reaches steady state. An innovative index, which represents the normalized kinetic energy gained by generators immediately after the fault, is used to minimize the search set of possible tripped lines. For every candidate line, the generator output powers are calculated considering the line as tripped, and the minimum error between the measured and the calculated generator powers is used to finally identify the tripped line. The IEEE 39-bus system and a 246-bus North Indian Grid representation are used to show the effectiveness of the method.