Historically, categorisation of customers has been made by fuse size, industry-code, type of residence, estimated annual consumption or actual consumption pattern. Through the hourly metering reform in Sweden, several calendar years of consumption data with hourly resolution are available, which enables new categorisation methods. Based on customers' hourly consumption, a data-driven methodology can be used to increase the knowledge about both individual customers and portfolios of customers. By doing so, assumptions in the prognosis-work can be avoided, retailers can develop deals tailored for individual customers and smart grid-components can more easily be identified. This study uses hourly metering data from approximately 150 000 customers to “cluster” these into categories. By calculating KPI's (Key Performance Indicators) from consumption profiles, customers can be grouped by how their KPI-values resemble each other using the method K-means. Nine KPI's are analysed to reflect several nuances in customers' consumptions, which enables a categorisation of customers with similar consumption patterns. The study shows that many customers in each segment resemble each other, but also that a significant share more distinctive traits. This substantiates a questioning of the methods used to define profiled customers today since these often assume a customer category to be homogenous.
No takes yet. Share an insight, caveat, or question.
Lindén et al. (2017) studied this question.