Conyac で依頼された翻訳結果を公開

Appliance load monitoring systems aim to achieve per appli- ance energy decomposition. Such systems however lack automated set-up, which precludes widespread roll-outs. This work illustrates how human supervision can be reduced to a strict minimum; appliance activity states are captured with cheap wireless sensors, which enables accurate automated energy data annotation. This paper demonstrates the accu- racy of this technique via an intuitive graphical user interface.Behavior altering technologies are being rolled out to better inform domestic energy consumers. Invariably, these manifest themselves as single devices offering direct feedback on energy expenditures,
software tools for comprehensive overview on energy spending, or complete systems for residential energy management. Such systems do not as yet incorporate realtime per-appliance energy decomposition. The recording of appliances’ signatures, the generation of annotated data sets to train machine learning algorithms, as well as the comparison of the systems’ outputs to the actual appliances’ activity all require manual input from homeowners.
This work demonstrates via an intuitive graphical user interface how appliances’ activity states can be autonomously retrieved using cheap wireless sensors and furthermore how this enables full automation of the appliance load monitoring set-up procedure. Using a temporary wireless sensor network to automate energy data annotation and configure a load monitoring system requires high accuracy. Kim et al. proved with Viridiscope the feasibility of indirect sensing to evaluate appliances’ power consumption with less than 10% error. Within this work, different combinations of cheap sensors and reasoning about concurrent sensor activity is proposed. Preliminary results showed an accuracy greater than 90% in laboratory experiments as our detection is confined to appliances’ activity; no additional errors from evaluating the power consumption are introduced.