how to use OR to prepare for weather-related power outages

I was pleased to find a Baltimore Sun article about Seth Guikema’s research on forecasting when and where major power outages will occur during storms like Hurricane Irene and this summer’s derechos. He has developed models that use weather forecast data and U.S. Census records for estimating power outages in regions affected by storms. I’ve written a couple of papers with Seth about adversarial risk analysis models for homeland security, but most of his work is in the area of natural disasters — and it’s really interesting. Seth has used data mining, generalized additive models, and hybrid data-mining/regression models to forecast power outages. He has found that soil and elevation is important for making accurate forecasts for power outages associated with hurricanes. The Baltimore Sun article notes that “In some areas, like New York City, variables like buried power lines made it less accurate. In others, like Rhode Island, detailed data on outages after the fact are difficult to obtain, making it hard to evaluate accuracy.”

Read the Baltimore sun article here.