Cavalieri et al. (2014) | CACAIE, 29(8), 590-607 | post-print (final draft post-refereeing) 1 Models for seismic vulnerability analysis of power net- works: comparative assessment Francesco Cavalieri, Paolo Franchin* Department of Structural and Geotechnical Engineering, University of Rome, La Sapienza, Via Antonio Gramsci 53, 00197, Rome, Italy & Jessica A. M. Buriticá Cortés, Solomon Tesfamariam School of Engineering, The University of British Columbia, Kelowna, BC, Canada Abstract: Electric power networks are spatially distrib- uted systems, subject to different magnitude and recurrence of earthquakes, that play a fundamental role in the well- being and safety of communities. Therefore, identification of critical components is of paramount importance in retro- fit prioritization. This paper presents a comparison of five seismic performance assessment models (M1 to M5) of increasing complexity. The first two models (M1 and M2) approach the problem from a connectivity perspective, while the last three (M3 to M5) consider also power flow analysis. To illustrate the utility of the five models, the well- known IEEE-118 test case, assumed to be located in central USA, is considered. Performances of the five models are compared using both system-level and component-level measures. Spearman rank correlation ρ is computed be- tween results of each model. Highest ρ values, at both sys- tem- and component-level, are obtained, as expected, be- tween M1 and M2, and within models M3 to M5. The ρ values between component-level measures are relatively high across all models, indicating that simpler ones (M1 and M2) are appropriate for vulnerability assessment and retrofit prioritization. The complex flow-based models (M3 to M5) are suitable if actual performance of the systems is desired, as it is the case when the power network is consid- ered within a larger set of interconnected infrastructural systems. 1 INTRODUCTION Critical infrastructure systems, such as power, water, transportation and communication networks, influence the well-being of a community. These systems are often ex- posed to low-probability high-consequence events (e.g., hurricanes, ice storms, tornadoes, earthquakes) that warrant a risk assessment. Interruption in any of these infrastruc- tures impacts quality of life, and can potentially result in economic losses and casualties (U.S.-Canada Power System Outage Task Force, 2004). The focus of this paper is on the seismic vulnerability as- sessment of electric power transmission networks (EPNs). Under different earthquake intensities, quantitative (proba- bilistic) information on the likelihood of damage, conse- quence of failure and extent of affected areas is valuable. Statistical models, based on large databases of outages from different hazards, are available in the literature. Examples of such models are given by the Negative Binomial Distri- bution, the Generalized Poisson Distribution and the Expo- nentially Accelerated Cascading Model. They provide for instance the spatial density of outages in a region (Liu et al., 2005), or the probability of cascading outages on transmis- sion lines (Chen et al., 2006). This information can be used to determine the required upgrading of an existing system, which can significantly reduce losses (e.g. Eidinger and Tang, 2012; Matsuda et al., 1991; Shumuta, 2007). Similar- ly, this information is also necessary for emergency plan- ning and post disaster recovery (including restoration of power) (e.g. Opricovic and Tzeng, 2002). This paper ex- plores instead physically-based models of seismic perfor- mance. An EPN is a spatially distributed and redundant system, whose physical behavior can be represented in terms of topological structure and operational state (Bompard et al., 2011). Thus, the performance analysis can be carried out at two different levels. The first level focuses on topological analysis of the system (i.e. connectivity analysis). Connec-