Problem Reformulation for Improved Regression Analysis in Granular Damage Forecasting

To facilitate the service restoration following outages due to hazardous weather conditions, accurate damage information, i.e., when and where the outage-causing damage occurs, is critical. Such information relies on high-resolution (spatial and temporal) outage estimation or prediction, which can only be enabled by developing models using granular, in space and time, data. In this study, to take full advantage of the high spatial and temporal resolution of utility outage data and weather radar observations, the regression problem for developing damage forecasting models is reformulated to consider the time evolving weather condition associated with each outage event. The reformulated problem considers the impacts on outages of variations and cumulative effects of weather conditions not only across the utility's service territory but also over time through the evolution of the storm. Using this reformulated approach, historical utility outage data are used to develop a new and improved damage forecasting algorithm and validate its performance improvement and applicability for more granular outage estimation.