Network-Informed Numerical Characterization of DC Network Static Voltage Stability Region

A static voltage stability region (SVSR) of a DC network is a set of load power injection profilesmodeled as power flow parameters. Any realization within the set ensures power flow equation solvability, i.e. the existence of a voltage-stable operating point. Accurately characterizing the SVSR is crucial for understanding DC network stability. Numerical methods have been developed that sample a large number of power flow parameters to find representative loading scenarios bearing SVSR geometry information. Simulation experiments are performed for each sample to determine its SVSR containment label, which manifests difficulties in keeping scalability with DC networks' growing sizes. To address this issue, we propose a network-informed numerical approach that uses DC network solvability conditions to reduce the computational workload by sample filtering and label extrapolation. First, we develop an algorithm that generalizes the SVSR-containment label of a single sample to that of an extended set to avoid dense simulations. Then, we construct an SVSR subset from the extended sets. The volume of the subset is expanded by an iterative closed-loop algorithm. In each iteration, we actively sample a small group of new parameters to label, and the revealed parameter space geometry information in turn is leveraged by a sample recommendation approach to filter samples carrying the most uncovered information for numerical labeling in future iterations. The newly labeled samples enlarge the SVSR subset, which aids in extrapolating the labels of the new samples. Case studies based on IEEE benchmark systems ranging from 14-bus to 1951-bus networks show that the proposed approach effectively reduces computational efforts while ensuring an SVSR subset.