Critical infrastructure systems such as water and electric power networks provide essential services that underlie the economic prosperity, security, and public health of the U.S. These complex, interdependent systems are prone to failure during hurricanes. Improved modeling of the ability of these systems to meet the needs of society after a hurricane makes landfall would substantially improve our ability to manage the risk of these systems failing. However, there are fundamental research needs of both conceptual and computational natures in the area of risk analysis for critical infrastructure systems in hurricane-prone areas. Conceptually, we do not yet have modeling frameworks that allow for accurate prediction of the performance of large-scale interdependent infrastructure systems during hurricanes, a necessary starting point for accurate risk assessment and management. Computationally, many of the available tools that aim to model infrastructure performance at the scale of large metropolitan areas require long run times on large computer clusters, limiting their usefulness for practical infrastructure planning and management. Recent advances in both statistical methods and computing based on graphical processing units (?graphics cards?) enable advances that can address both the conceptual and computational limitations inherent in current approaches for risk analysis for interdependent infrastructure systems in hurricane-prone areas. The focus of this project is on developing methods for accurate performance and risk modeling for interdependent infrastructure systems, methods that are practical for infrastructure managers to use. While the focus of this project is on coupled water and power systems, the advances will have application much more broadly.
This project will enable significantly more accurate and rapid risk analysis for interdependent infrastructure systems, allowing highly limited public infrastructure funds to be spent more efficiently and helping to better protect economic and public health during disasters. The models developed in this project will be practical for use on desktop computers with existing higher-end graphics cards, greatly enhancing the ability of infrastructure managers to run these models on their existing computer hardware. In addition, this project will yield insights into the factors that lead interdependent infrastructure systems to be more resilient during a hurricane, helping engineers and utility system managers better understand how to strengthen their systems. In parallel with the research efforts, this project will aim to interest students traditionally underrepresented in engineering programs in pursuing engineering as a career. This will be done through outreach at multiple educational levels.
This project developed improved methods for estimating the robustness and reliability of infrastructure systems impacted by severe external events. These methods allow power outages to be predicted accurately in advance of an approaching hurricane as seen in the figure showing predictions of power outages for Hurricane Irene. Research from this project also gives a better understanding of when detailed engineering models are needed to accurately estimate the reliability of power systems and when simpler topology-based models are sufficient as well as how a network’s resistance to failure is related to its topology – the arrangement of the nodes and links in the network. Together, these advances allow for more rapid, less computationally demanding estimation of infrastructure reliability. The other main advance in the project was to develop a method for estimating infrastructure performance at a number of different spatial scales on the basis of a spatially detailed performance model. This allows the output from complex infrastructure models to be communicated at the scales most natural to decision-makers such as utility mangers and local, state, and federal emergency response managers. This rescaling is done in a way that is fully consistent with the underlying detailed models and that can convey the uncertainty in the underlying estimates. The advances in this project provide tools for those managing utilities or dependent on utilities to better predict and prepare for the impacts of significant events such as hurricanes. Indeed, the hurricane power outage forecasting model further developed in this project is already in operational use in the electric power industry. This project also contributed to the development of future researchers in the fields of infrastructure engineering and risk analysis. Two female Ph.D. students and one female postdoctoral researcher were involved in this project, gaining experience and expertise in infrastructure modeling and risk analysis. These students and the postdoctoral researcher intend to pursue research careers in this area in the U.S., helping to secure America’s infrastructure and reduce the impacts of future disasters.
