Application of clustering for electricity usage patterns detection

Abstract

Electricity usage patterns are important for suppliers in order to ensure efficient electricity distribution and pricing, and for users to save costs. Smart meters provide high granularity electricity usage data (e.g., hourly), which can be used to extract patterns of daily/weekly/monthly/ yearly electricity usage. In this research we propose a method for grouping electricity users based on hourly electricity usage data, based on the k-means clustering and periodicity analysis, with semi-automatic parameters selection based on adequacy measure. We illustrate the method with 1500 electricity users 3 years’ data, discuss pros and cons of the methods, and future plans. Acknowledgements We thank Binar Solutions for cooperation and European Social Fund and the Republic of Lithuania (grant number VP1-3.2-ŠMM-01-K-02-002) for funding.