Thursday, 11 January 2018: 10:45 AM
Room 15 (ACC) (Austin, Texas)
Weather events are a major cause of power outages in the United States. Weather related power outages cost the U.S. an estimated 18 to 33 billion dollars annually in economic losses. Much of the previous research on predicting power outages has focused on hurricane related outages. However, on average, hurricanes cause less than half the number of customer outages as thunderstorms. This research utilizes machine learning techniques on a large data set consisting of historical outage data and weather forecasts combined with environmental data such as population density, land cover, soil, and tree characteristics to develop a thunderstorm outage prediction model. This research also evaluates how changes in the spatio-temporal resolution of the model affect its predictive ability. This gives insight into the spatio-temporal resolution that will maximize the usefulness of the model when used in an operational setting.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner