Abstract

Critical infrastructures (CIs) are spatially distributed infrastructure networks that supply essential services and goods to our society. However, CIs are exposed to diverse disruptive and severe natural events, random failures and manmade attacks, which undermine their proper functioning producing negative impacts on the population. In this context, it is of paramount importance to identify which are the most relevant CI components that deserve special attention for improving CIs protection. The identification of critical components is a well consolidated practice in risk analysis of complex technological systems; however, when looking at real-world spatially-distributed CIs, identifying the most critical components precisely is challenging. Indeed, CIs cover large spatial areas, have variable degrees of redundancy, and often exhibit irregular topological characteristics. In order to address this problem, this work looks at the ability of different graph theoretical measures or scores at identifying critical components in power distribution networks to later compare them. The scores considered herein include various network centrality measures (degree, eigenvector, closeness, betweenness) and some variations. The comparison of these measures is performed on three power distribution networks from central Chile, finding that a variant of betweenness centrality is the best at identifying the most critical components.