Speaker :Ms. M. Kalyani (EE11D007)
With the increased interconnections in restructured environment, power systems are being operated close to the stability limits for several economic reasons. Stressed system conditions together with inadequate situational diagnostic support may cause catastrophic failures leading to a blackout. The islands formed automatically after a sequence of cascading events, such as line tripping usually suffer from large mismatch in load-generation. So, the uncontrolled islands are highly prone to blackouts. The study of historic blackouts reveal that most blackouts are due to the maloperation of the local relays resulting in uncontrolled islands with excessive electrical imbalance and if proper system splitting with load-generation balance had been performed in time many blackouts would have been avoided. So, when there is a loss of synchronisms of generators and emergency control actions cannot keep the integrity of the power network, controlled islanding is considered as an effective defense strategy to prevent the system from catastrophic blackouts. Controlled islanding is generally solved either as a constrained combinatorial optimization problem or a slow coherency based linearized approach. The combinatorial explosion of the solution space of a large power network increases the complexity of solving, while the linearized slow coherency approach cannot track the varying coherent generator groups with change in network operating conditions. The present study proposes a novel slow coherency based network splitting technique that groups generators and load buses simultaneously from the measured signals, ensuring generator coherency. The coherent bus groups are determined form the slow modes of bus voltage signals using Zolotarev polynomial based filter bank (ZPBFB). The dimensional reduction techniques are used to cluster the slow coherent buses and the islanding boundaries are determined. The proposed method is demonstrated on IEEE 39-bus and 118- bus test systems and the effectiveness of the proposed method is compared with graph spectra based controlled islanding.