Discovering routine behaviours in smart water meter data

Abstract

Smart water meters are being used on a large scale by water providers to record hourly water use of households. The time series data recorded by smart water meters provide real-time information about water use activities. This paper proposes an algorithm to automatically discover recurrent routine behaviours in smart water meter data. The recurrent routine behaviours characterize regular water use activities during consecutive hours, which occur multiple times in a period. Our algorithm differs from previous exact motif discovery algorithms because we discover frequently occurring short subsequences with variable length. Experiment on a real-world dataset collected from an inland town of Kalgoorlie-Boulder in Western Australia demonstrates that the proposed algorithm discovers useful recurrent routine behaviours of different lengths, which are relevant for domain experts.