A Mobile System for Annotation of Home Energy Data

Home energy management is increasingly important. Though there are a plethora of tools for aiding home energy management, few provide concrete suggestions for helping users to manage energy demand. A key component in developing automated energy management schemes is drawing a connection between energy usage and a user’s context. Existing approaches either rely on users to annotate data after the fact, or rely on intrusive and costly sensing systems deployed in the home. This work presents a system for collecting in situ annotations using a mobile application coupled with an existing home energy measurement infrastructure. We use a novel power profiling approach to determine when appliances transition from an idle to active state and aggressively prompt users to provide annotations of their current context. In a five-week study, we were able to collect an average of over 2 annotations per day and users provided a wide range of annotations, however the overall response rate was lower than expected leading to a sparse data set. We conclude by examining the utility of sparse energy data annotations. We first demonstrate that our data set confirms that end user annotation is necessary—a completely automated activity inferencing scheme is implausible. We further demonstrate that sparse annotation data can be used to predict a user’s activity with an accuracy of more than 50% in hours where our data set contains an annotation from the user. We finally consider the feasibility of using annotations to predict a user’s energy needs.