Discussion about the Weather Impact on the Daily Outages in Urban Distribution System

Abstract

In this paper, an evaluation approach for analyzing the weather’s impact on the number of daily outages in the urban distribution system is explored. By dividing the number of outages into two levels, the task could be carried out as a binary classification problem. In this study, the actual outage data from the distribution system operator is analyzed together with the local weather condition records. First, the tendency of different outage levels to weather conditions is described by the Principal Component Analysis (PCA). Then, the Support Vector Machine (SVM) algorithm is adopted to build the classification model for predicting the outage levels based on the weather condition. An oversampling method is introduced to manage the severe imbalance between the two outage levels. At the end, the performance of the classification model is assessed with the Receiver Operating Characteristic (ROC) curve.