<p dir="ltr">A broad scope of this dissertation is to verify that a nearby loss of generation event in power system can be distinguished from similar remote disturbances by analyzing the resulting local modes of oscillation. An oscillation-based index derived from methods like Fourier transform, sinc filters and resonant filters is devised and experimented in combination with a variant of df/dt index to jointly classify if a loss of generation event is nearby or remote. A phenomenon widely observed during a loss of generation event is the average decrease in the system’s frequency, typically monitored using the df/dt index. Under-frequency load-shedding (UFLS) relays that are based on df/dt are highly likely to trip for nearby frequency events when combined with the oscillation-based index we propose. Nearby in our context refers to geographical distance, which is correlated with electrical distance, and includes buses within about 50-100 miles of the event location.</p>

Severe disturbances in a large electrical interconnection cause a large mismatch in generation and load in the network, leading to frequency instability. If the mismatch is not rectified quickly, the system may disintegrate into multiple islands. Though the Automatic Generation Controls (AGC) perform well in correcting frequency deviation over a period of minutes, they are ineffective during a rolling blackout. While traditional Under Frequency Load Shedding Schemes (UFLS) perform quick control actions to arrest frequency decline in an islanded network, they are not designed to prevent unplanned islanding. The proposed Intelligent Load Shedding algorithm combines the effectiveness of AGC Scheme by observing tie line flows and the speed of operation of the UFLS Scheme by shedding loads intelligently, to preserve system integrity in the event of an evolving cascading failure. The proposed scheme detects and estimates the size of an event by monitoring the tie lines of a control area using Wide Area Measurement Systems (WAMS) and initiates load shedding by removing loads whose locations are optimally determined by a sensitivity analysis. The amount and location of the load shedding depends on the location and size of the initiating event, making the proposed algorithm adaptive and selective. Case Studies have been presented to show that control actions of the proposed scheme can directly mitigate a cascading blackout. / Ph. D.