Self-configuring event detection in electricity monitoring for human-building interaction

摘要: Abstract Monitoring the temporal changes in operational states of appliances is a key step inferring dynamics operations smart homes. This information could be leveraged variety energy management applications including breakdown individual loads, occupancy patterns, and associating use to occupants’ activities. The identified through detecting classifying events on power time-series. Despite advancements field event detection, they often require in-situ configuration model parameters achieve higher level performance according each new context. In order address such limitation, this paper, we have proposed self-configuring detection framework for appliances. seeks autonomously learn contextual characteristics loads from environment adapt parameters. unsupervised couples an automated clustering identifying recurring motifs, which are representations appliances’ transient draw signatures given proximity-based motif matching events. was evaluated EMBED dataset, publicly available fully labeled electricity disaggregation collected three apartments with different categories evaluations demonstrate that outperforms conventional classes across environments. also facilitate human-building interactions training home by populating motifs infer activities occupants.