Identifying water consumption profiles through unsupervised clustering of household timeseries: the case of Attica, Greece

Urban water systems are complex, socio-technical systems, tasked with meeting water demands, generated through a continuous (socio-technical) interplay between customers and infrastructure, maintaining high reliability levels. Understanding demand patterns at different scales, is thus essential to manage the associated water distribution infrastructure and better serve customers. Here, we analyse water consumption data from 40 municipalities in Attica, Greece, served by the water company of Athens, EYDAP S.A., using machine learning techniques to detect principal patterns in water consumption. The data, which were extracted and provided by EYDAP S.A., are monthly time series of consumption points between 2010 and 2021, a sample which is analysed through an unsupervised data clustering process, to gain insights on dominant consumption patterns and to identify characteristic customer profiles.