Cyclone risk assessment of large-scale distributed infrastructure systems

Abstract

Coastal communities are vulnerable to tropical cyclones. Community resilience assessment for hazard mitigation planning demands a whole-of-community approach to risk assessment under tropical cyclones. Community risk assessment is complicated since it must capture the spatial correlation among individual facilities due to similar demands placed by a cyclone event and similar infrastructure capacities due to common engineering practices. However, the impact of such spatial correlation has seldom been considered in cyclone risk assessment.

This study develops advanced stochastic models and methodology to evaluate the collective risk of large-scale distributed infrastructure systems under a scenario tropical cyclone, considering the spatial correlations of wind demands and structural capacities modelled by fragility functions. Wind-dependent correlation of fragility functions is derived from the correlation of structural resistances using joint fragility analysis. A general probabilistic framework is proposed to evaluate the damage of infrastructure systems based on joint fragility functions, where the stochastic dependence between the fragility functions of individual facilities is approximated by a Gaussian copula. A stochastic model is developed to model the spatially correlated wind speeds from a tropical cyclone, when wind speed statistics based on three cyclone wind field models of different complexity are examined. The impact of wind speed uncertainty and spatial correlation on risk assessment is investigated by evaluating the cyclone loss of an electric power system, when three loss metrics are examined including damage ratio, power outage ratio and outage cost to electricity customers. Since the risk assessment of a large-scale infrastructure system is computationally challenging, an interpolation technique based on random field discretization is developed, which can simulate spatially correlated damage to infrastructure components in a scalable manner.