2019 CSCE Annual Conference - Laval (Greater Montreal)

2019 CSCE Annual Conference - Laval (Greater Montreal) Conference


Title
EVALUATING NATURAL HAZARD-INDUCED ELECTRICITY SECTOR INOPERABILITY LEVERAGING DATA-DRIVEN STATISTICAL LEARNING APPROACH


Author(s)
Mrs. Soojin Yoon, Construction Engineering and Management, Purdue University (Presenter)
Dr. Sayanti Mukherjee, University at Buffalo
Dr. Makarand Hastak, Division of Construction Engineering and Management, Purdue University
Abstract

Our built environment is threatened with ever-increasing risks of climate change and natural hazard-induced extreme events. Under multi-hazard scenarios—such as floods, wildfires, droughts, heatwaves, etc.—the risk of severe weather-induced power outages is growing. Such weather–induced cascading power outages can cause tens of billions of dollars of economic loss. In this research, we propose an integrated approach to develop a composite predictive model framework for evaluating the inoperability of the utility sector in a multi-hazard scenario. Inoperability is defined as the level of utility sector’s service disruption (i.e., interruption to the power supply measured by power outage duration) caused by the severe weather events. The proposed framework consists of three main phases: (1) to select natural hazard-induced power outage events, (2) to develop separate models for each of the disaster categories as discussed before, and (3) to develop a composite predictive inoperability of the utility model by integrating the individual models developed in Phase 2. We leveraged data-driven semiparametric statistical learning model called generalized additive model (GAM) to develop the inoperability prediction model. The data on major power outage events in the continental U.S. between 2000 and 2016 is used to develop the model. Our proposed inoperability prediction model will allow the stakeholders and decision makers (e.g., the state regulatory commissions or the utility companies) to evaluate the extents of electricity sector inoperability, both under a single-hazard or a multi-hazard scenario. The model results will help in risk-informed decision-making for pre- and post- disaster resource allocation and capacity expansion investments to improve the security of the electricity sector as a whole.