Distribution Network Outage Data Analysis and Repair Time Prediction Using Deep Learning

This paper presents a statistical analysis for power outages in the distribution network. 6-year outage data in the service territory of one U.S. utility is used in the analysis, which includes several storms and hurricanes. The first part of this paper summarizes the outage data with a focus on weather event-related outages; the second part proposes a machine learning-based technique to predict the repair and restoration time after outages, which provides useful information for utilities' repair work scheduling and crew dispatching. A Deep Neural Network (DNN) is trained using supervised learning with 5-year outage data. The 6th year data is used to validate the performance of the trained DNN model.