Search

A well-designed scheme is essential for assessment and protection of the stability of power system after an unanticipated event. This paper proposes a strategy for regulating three-phase faults and other events that endanger frequency or voltage stability of the system. The aim is to promptly identify such events utilizing supervised learning-based classification modules, and initiate suitable emergency control. The imbalances of active and reactive power after an event have been utilized to predict the stability of system. A new optimization-based formulation has been presented for estimation of reactive power imbalance. An event predicted as unstable is counteracted by prompt load shedding. Conventionally, load shedding is based on frequency or voltage information independently which reduces the effect of load shedding. This work proposes a new load shedding procedure considering both active and reactive power. The proposed scheme is tested on New England (NE) 39-bus system and Northern Regional Power Grid (NRPG) 246-bus system, and results of load shedding have been compared with two existing load shedding schemes. The results indicate the accuracy of classification modules in identifying faults and unstable events, and the proposed load shedding restores stability of the system with less amount of load shed and better voltage profile. [ABSTRACT FROM AUTHOR]

This article presents a real-time event diagnosis approach for improved situational awareness in power systems using wide area synchrophasor measurements. A new index for early detection of events and a scheme for their geographical localization is proposed. A supervised learning-based event classification module is proposed for real-time identification of the events. The discontinuities generated in voltage and frequency measurements due to events are adequately localized in the wavelet domain. Therefore, the discrete wavelet transform is used for detection and localization of events. Time-frequency features are extracted from voltage and frequency measurements, and a multi-class support vector machine classifier is trained. A linear relationship between wavelet coefficients and the magnitude of disturbance triggered by the event is obtained. The proposed work functions as a supervisory layer above the traditional protection equipment and can validate its operation. The effectiveness of the proposed approach is demonstrated on the standard IEEE New England 39 bus test system. [ABSTRACT FROM AUTHOR]