Non-intrusive disaggregation of water consumption data in a residential household

The water conservation campaigns in residential households are hindered by the poor understanding of residents of how much water they use. For the better designed interventions new tools are necessary to educate the consumers on the water usage of different consumption events. In this paper we use the fine grained (0.5 Hz) water consumption data that was collected nonintrusively in a household over the period of 21 days to develop such tools. We examine the collected data and disaggregate the consumption events into three different categories: short events (e.g., toilet flush), long regular events (e.g., washing machine) and long irregular events (e.g., showers). To achieve this, we use clustering methods, based on level set trees, to identify groups of events that are similar to each other.