A novel approach for instability detection based on wide-area measurements and new predictors

This paper proposed a novel approach based on the center of power (COP) applied on a sliding window for predicting real-time network instability. The COP speed reference and two new indices (γ and κ) derived from phasor measurements units serve as predictors in the building of random forest (RF) learning. This approach reduced the number of variables to be analyzed and allowed to decide the state of the network at any time without using fault detectors. Two study cases are carried out on the IEEE 4 generators buses and Australian 14 generators to validate the effectiveness and the performance of the proposed method.