A graph theoretic approach to accelerate natural cutset prediction during an out-of-step condition

Abstract

When an out-of-step (OOS) condition occurs in a power system, it splits into multiple islands. This uncontrolled separation happens due to distance relay operation on the lines where electrical centers appear. Real-time phasors obtained from Phasor Measurement Units (PMUs) can be used to track the evolution of electrical centers and identify the natural boundary of separation (cutset). Interestingly, electrical centers do not appear simultaneously on the cutset lines; instead, they evolve at different rates. Hence, the cutset is identified only when the signature of the last electrical center appears. We aim to accelerate the cutset prediction so that the complete cutset is identified when the first electrical center is predicted. For this, we propose a computationally efficient algorithm based on the application of graph theory and clustering technique. Thus, the proposed scheme addresses the significant delay in cutset prediction and thereby increases the time available for controlled is landing. Test results on standard IEEE systems and the All India test system validate the scheme.