Coherency-Constrained Spectral Clustering for Proactive Load Shedding: A Tri-Level Model

Abstract

In low-inertia power grids, traditional load-shedding methods can be ineffective, causing instability and unnecessary grid islanding. This work introduces a distributed trilevel optimization model for proactive load shedding to prevent grid islanding considering generator coherency identification. Preventing islanding is crucial because it reduces the need for additional load shedding and transmission line tripping, thereby maintaining frequency stability. This is especially important in systems heavily dependent on non-dispatchable resources like wind and solar. The model employs generator coherency identification through spectral clustering using synchronizing coefficients. The results are then used as constraints in a dynamic graph based on power flow measurements to identify coherent areas for load shedding under highly stressed conditions. This approach offers practical applications and holds promise for realworld use.