Daily load curves in distribution networks — Analysis of diversity and outlier detection

Abstract

This work is focused on diversity analysis of daily load curves and outlier curve detection in distribution networks. For diversity evaluation unified load curves are defined and the Euclidean distance between them and the average curve is calculated. Outlier detection is based on the distance between a load curve and an average curve representing curves of the same day of the week as the analyzed curve and from the same period of time. For detection of outliers in shape normalized daily curves are defined and the distance is measured between them and the average normalized curve. The analyzes are illustrated using real-world data from more than hundred substations.