Critical-cluster identification in transient stability studies

Abstract

A method is proposed for identifying the critical cluster of machines, i.e. the machines responsible for loss of synchronism in a power system following a large disturbance. It is based on the conjecture that the loss-of-synchronism condition can be recognised by, first, considering ‘near-critically cleared trajectories’ of the machines and, secondly, observing how they are organised near the system's unstable equilibrium point. The term ‘near-critically cleared trajectories’ is meant to imply the swing curves of the system in the postfault-phase and is computed for a fault-clearing time that is slightly larger than the actual critical clearing time. To realise the above, the method uses the extended equal-area criterion. This direct criterion makes it possible to assess a convenient clearing time for computing the swing curves; to determine the system's unstable equilibrium point with great ease; and to select the critical machines by observing them at the time corresponding to this unstable equilibrium point. It also makes the speed of critical-cluster identification compatible with real-time requirements. Examples are given using the IEEE test system to illustrate the essential features: reliability in correctly identifying the critical clusters, robustness with respect to its capacity to do so under very stringent conditions, and effectiveness concerning its ability automatically to identify critical clusters of any size, i.e. irrespective of the number of machines that they contain.