Model order selection for probing-based power system mode estimation

The paper discusses model order selection for probing mode estimation algorithms. Four methods are analyzed and compared: 1) Residual analysis based model order selection, 2) Model order selection based on singular values, 3) Akaike Information Criterion, and 4) Variance-Accounted-For (VAF) as a measure of optimal fitting between measured data and model response. The methods are assessed using synthetic PMU measurements from the simulation of the KTH Nordic 32 Test system and the IEEE test system with 50 generators and 145 buses.