A data-driven analysis of outage duration in power distribution systems

This paper presents a data-driven approach for exploring outage duration in power distribution systems. The primary goals are to analyze and interpret the outage duration according to the underlying causes and to identify significant variables that strongly impact the outage duration. To carry out this study, actual outage data collected by a major utility company that has operations in the southeastern US is utilized. First, practical issues associated with the outage data set are discussed, and necessary data pre-processing tasks are carried out. Afterward, two types of analysis, namely statistical and visual analytics are conducted to investigate several features that could contribute to outage durations. Also, field engineering insights are provided to interpret the quantitative results. Finally, the importance of all features is examined by employing a data analytics algorithm and the most important features are identified. The results of this paper could be used by utilities to enhance their outage management systems and to gain valuable information for carrying out essential preventive maintenance for the purpose of reducing the duration of future outages.