Algorithms for Detecting Nearby Loss of Generation Events for Decentralized Controls

Abstract

The paper describes algorithms to screen realtime frequency data for detecting nearby loss of generation events. Results from Fourier calculation are combined with other features to effectively distinguish a nearby loss of generation from similar remote disturbances. Nearby in this context usually refers to an event occurring around 50-100 miles from the measurement location. The proposed algorithm can be trained using pattern recognition tools like decision trees to enable smart devices including appliances like residential air conditioners and dryers to autonomously detect and estimate the source of large frequency disturbances. An area of application of this strategy is to actuate controls such as location targeted under frequency load shedding (UFLS) so that loads closest to a tripped generator are the most likely to shut down.