Comparing dynamics of cascading failures between network-centric and power flow models

Small initial failures in power grids can lead to large cascading failures, which have significant economic and social costs, and it is imperative to understand the dynamics of these systems and to mitigate risks of such failures. We directly compare two prominent models for cascading failures in power grids: a power flow model-ORNL-PSerc-Alaska (OPA) and a complex networks model-Crucitti-Latora-Marchiori (CLM). Quantitative comparison of these two models is not trivial and has not been previously performed, so we present a method to quantitatively compare the two using transmission capacity as a common variable. Primarily, we find that the two models exhibit similar phase transitions in average network damage (load shed/demand in OPA, path damage in CLM) with respect to transmission capacity, sharing a common critical region and similar transitions in probability distributions of network damage size. Furthermore, we find that both OPA and CLM reveal similar impacts of network topology, size and heterogeneity of transmission capacity, with respect to vulnerability to large cascades. Thus, our analysis indicates that the CLM model, despite neglecting realistic power flow assumptions and exhibiting differences in behaviour at the local scale, nonetheless exhibits ensemble properties which are consistent with the more realistic OPA fast-scale model. Given the advantages of simplicity and scalability of CLM, these results provide impetus for the use of complex networks models to study ensemble properties of cascading failures in larger power grid networks.