Disaggregated water sensing from a single, pressure-based sensor: An extended analysis of HydroSense using staged experiments

We present an extended analysis of our previous work on the HydroSense technology, which is a low-cost and easily installed single-point sensor of pressure for automatically disaggregating water usage activities in the home (Froehlich et al., 2009 [53]). We expand upon this work by providing a survey of existing and emerging water disaggregation techniques, a more comprehensive description of the theory of operation behind our approach, and an expanded analysis section that includes hot versus cold water valve usage classification and a comparison between two classification approaches: the template-based matching scheme used in Froehlich et al. (2009) [53] and a new stochastic approach using a Hidden Markov Model. We show that both are successful in identifying valve- and fixture-level water events with greater than 90% accuracies. We conclude with a discussion of the limitations in our experimental methodology and open problems going forward.