Improved Statistical Modeling of Power Outages During Hurricanes in the United States

• Activity Number: 307
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Physical and Engineering Sciences
• Author(s): Roshi Nateghi*+ and Seth Guikema and Steven Quiring
• Companies: The Johns Hopkins University and The Johns Hopkins University and Texas A&M University
• Keywords: power outage ; classification trees ; Generalized Additive Model

Abstract

Hurricanes frequently cause widespread and prolonged power outages in the United States that affect our security systems, digital economy and millions of customers. Accurate outage forecasts before a storm would allow utilities to better plan their recovery efforts and the public and public agencies to better plan their responses. Previous work by the authors has developed accurate forecasting models on a hybrid Classification Tree - Generalized Additive Model framework. This paper extends this approach by including substantially more detailed data about system maintenance and tree trimming activities near power lines. The models utilizing this more detailed data are compared with prior models through a hold-out cross-validation study. Issues of practical implementation within utility company emergency preparation and response groups are also discussed.